epic: backlog burndown -- tracking all open work

[bpowers]

Purpose

A single, pinned index of all known open work, so anyone (human or agent) can find a bite-sized thing to pick up without re-surveying the whole tracker. This is the backlog-burndown drive: chip away, tick boxes, keep the count falling.

This is a master index. Two large clusters have their own dedicated trackers and are not re-listed here:

• Loops that Matter (LTM) -- 35 open issues curated in epic: Loops that Matter (LTM) -- tracking known issues and improvements #488 (pinned). Filter: gh issue list --label ltm --state open.
• Dormant feature epics -- big-rock features that are designs unto themselves: array support #25 (array support), unit checking #35 (unit checking), module editing #46 (module editing), fork and merge models #57 (fork & merge models).

Everything else is grouped below by code area. Each row links a standalone issue.

How to use this tracking issue

• Looking for something to pick up? Browse a section that matches your interest and grab an unchecked box. bug/hygiene items tend to be self-contained; enhancement items are larger.
• Filtering on the command line: every issue now carries area labels -- gh issue list --label engine --state open, --label frontend, --label backend, --label security, --label layout, --label ci, --label docs, --label mcp, --label compat, --label hygiene.
• Closed something? Tick the box here in the same PR (or a fast-follow). The box is the human-readable progress signal; the underlying issue state is the source of truth.
• Filed something new? Add a - [ ] #N -- short title row to the right section (or to epic: Loops that Matter (LTM) -- tracking known issues and improvements #488 if it is LTM).

Recently triaged (2026-07-28)

• Full sweep of every open issue against this index. 71 untracked issues were added. The largest new cluster is the Stella conveyor/queue engine (16 issues), which now has its own section; the second largest is build/lint tooling.
• Ticked 11 rows whose issues had been closed since the last sweep: AddressSanitizer: resvg 0.47.0 type mismatch on nightly Rust #287, engine: two divergent circular-dependency cycle-gate code paths (System A salsa vs System B model.rs:442) #568, engine: arrayed VECTOR SORT ORDER returns flat global indices instead of per-iterated-slice 0-based ranks #585, server: @simlin/server Jest suite intermittently fails one test under parallel pre-commit load #635, engine: salsa::Update on compiled bytecode uses PartialEq over f64/NaN (benign missed-backdate) #642, embed: cross-origin web component broken -- module worker chunk fetched from app.simlin.com with no CORS/fallback #688, server: preview rendering lost worker-thread isolation; no timeout + unbounded GAE instances compound the DoS/cost surface #694, deploy: .gcloudignore doesn't import .gitignore; gcloud app deploy uploads untracked files and trips GAE 10k-file cap #695, Canvas view changes trigger a full engine round-trip per frame during pan/zoom #707, rsbuild/rspack tree-shakes reset.css side-effect import out of src/app production CSS bundle #708, diagram: component-library CSS modules hardcode light colors with no dark-mode overrides #709.
• The old "Deploy / ops / CI / build / test hygiene / CLI" bucket had grown to ~40 rows, which defeats the browse-and-pick purpose. Split into Build / lint / type-check tooling, CI / test hygiene, and Deploy / ops / release / CLI.
• LTM-labeled issues ltm: every-emitted-fragment-compiles harness over a generated corpus (prevention) #782, ltm: link-score generation materializes one full guard-form arm per element of an arrayed target (quadratic in element count) #977 and ltm: the ceteris-paribus wrap leaves a LOOKUP table argument's runtime index live in every link-score partial #984 were routed to epic: Loops that Matter (LTM) -- tracking known issues and improvements #488 per the convention below (and ltm-finding: discovery mode mints phantom from-side element keys for Bare A2A edges (mapped-dimension and scalar-source legs; loops undiscoverable) #754/ltm: de-hoisted array-valued RANK gets diagonal-only element edges, so rank-mediated cross-element loops are not enumerated #776 ticked there), so they are deliberately absent here.
• Cannot log in locally #78, Loading Stella models - some issues #90 and FR: Simlin-engine as separate crate #91 predate this epic and were missed by the original index; they are now filed under the matching sections.

Recently triaged (2026-06-08)

• Changing React code to use functional components #105 -- React class-to-function migration complete (only ErrorBoundary stays a class, by necessity); closed.
• logout does not work? #83 -- duplicate of Logout flow is unimplemented (client menu item is a no-op; DELETE /session is a non-responding stub) #524 (logout); closed.
• All 123 previously-unlabeled issues were given area labels; new labels layout, ci, docs, security, mcp were created.
• Needs your call: Full app local dev still requires Firebase (no lightweight/SQLite backend) #109 ("needs heavy libraries + Google account to run locally") is partially addressed by the @simlin/serve local viewer but the full app still needs Firebase -- left open pending a decision on whether to narrow or close it.

---

Engine -- Vensim / XMILE / MDL / systems-format compat & macros

• XMILE export from simlin: variable name to lowercase but not in equations #82 -- XMILE export lowercases variable names outside equations but not within
• Loading Stella models - some issues #90 -- Stella .stmx import: empty <variables/>, quoted hyphenated idents, connector-supplied external data
• User can proceed with importing invalid XMILE file despite an error #101 -- user can proceed past an error when importing an invalid XMILE file
• Add MDL round-trip integration test for special-character identifiers #335 -- add MDL round-trip integration test for special-character identifiers
• engine: GET DATA BETWEEN TIMES + implicit .dat loading blocks 2 simulation tests #347 -- GET DATA BETWEEN TIMES + implicit .dat loading (blocks 2 sim tests)
• engine: C-LEARN macro expansion blocks 1 simulation test #349 -- C-LEARN macro expansion (blocks 1 sim test)
• MDL writer does not reconstruct GET DIRECT calls for arrayed variables #393 -- MDL writer doesn't reconstruct GET DIRECT calls for arrayed variables
• Systems format translator: duplicate source/dest flow pairs produce colliding identifiers #399 -- systems format: duplicate source/dest flow pairs collide on identifier
• Systems format writer groups flows by source stock, breaking round-trip for cross-stock references #400 -- systems format writer groups flows by source stock, breaking round-trip
• MDL roundtrip loses original view coordinates due to compose_views offset #401 -- MDL roundtrip loses view coordinates (compose_views offset)
• MDL roundtrip doesn't preserve custom sketch display shapes #402 -- MDL roundtrip drops custom sketch display shapes
• MDL writer pipe connector flags may not match Vensim originals #403 -- MDL writer pipe-connector flags may not match Vensim originals
• MDL writer produces invalid output for projects with no views #404 -- MDL writer emits invalid output for projects with no views
• MDL roundtrip loses per-view font metadata in multi-view models #405 -- MDL roundtrip loses per-view font metadata in multi-view models
• Refactor write_view_segment in MDL writer to reduce parameter count #406 -- refactor write_view_segment to reduce parameter count
• Systems translator does not reject stocks named after engine-reserved identifiers #440 -- systems translator accepts stocks named after engine-reserved identifiers
• engine: test_mdl_equivalence fails on main -- native MDL parser emits 3 extra link elements for SIR.mdl #512 -- test_mdl_equivalence: native parser emits 3 extra link elements for SIR.mdl
• engine: Vensim macro formal parameter named like a control variable (Final Time, etc.) is silently dropped on MDL import #552 -- Vensim macro formal param named like a control var is silently dropped
• engine: single-parameter Vensim macro invocation is silently rewritten to LOOKUP on MDL import #553 -- single-parameter Vensim macro is silently rewritten to LOOKUP on import
• engine: macro whose body wraps a same-canonical-name intrinsic (INIT = INITIAL(x)) causes false recursion cycle, blocking C-LEARN macro expansion #554 -- macro INIT = INITIAL(x) false-recursion cycle (blocks C-LEARN expansion)
• engine: metasd COVID model not simulation-tier-eligible -- unsatisfied GET DIRECT/GET XLS DATA external data blocks full SSTATS reference validation (Phase 7 corpus scope) #555 -- metasd COVID model not sim-tier-eligible (unsatisfied GET DIRECT/GET XLS)
• engine: add proptest for XMILE macro round-trip and extract/reconstruct idempotence invariants #556 -- add proptest for XMILE macro round-trip / extract-reconstruct idempotence
• engine: XMILE macro-invocation argument recovery relies on undocumented Phase-4 ModuleReference emission order (latent silent-reorder bug) #557 -- XMILE macro arg recovery relies on undocumented emission order (latent bug)
• engine: C-LEARN full simulation blocked by non-macro issues (circular dep, dimension mismatches, unknown dep) -- the macro epic's deferred 'tracked separately' follow-up #559 -- C-LEARN full simulation blocked by non-macro issues
• engine: metasd macro corpus SIMULATION tier is empty -- enable it (Vensim DSS reference outputs starting with Theil_2011 + per-model non-macro blocker triage) #561 -- enable the metasd macro corpus SIMULATION tier
• engine: Variable::Module inside a macro-body model is silently dropped by the MDL writer (project_to_mdl reject gate scans only the main model) #566 -- Variable::Module inside a macro-body model dropped by the MDL writer
• engine: B2-A intentional flow-subscript divergence (eae79426) needs xmutil-gated mdl_equivalence/test_clearn_equivalence verified-or-updated before B2 PR (same class as #512) #570 -- verify B2-A intentional flow-subscript divergence vs xmutil goldens before PR
• C-LEARN: residual Ref.vdf numeric divergence after element-cycle-resolution (SAMPLE UNTIL / INIT-:NA: / numeric tail / NaN-vs-NA) #591 -- C-LEARN residual Ref.vdf numeric divergence (SAMPLE UNTIL / :NA: / tail)
• mdl: importer expands single apply-to-all Vensim equations into N identical per-element equations #651 -- MDL importer expands single apply-to-all equations into N identical per-element ones
• engine: 2022-era models with quoted literal-period idents silently fail to simulate (no load-time canonical-data migration) #690 -- 2022-era models with quoted literal-period idents silently fail to simulate
• engine: freshly imported/created period-named variables may fail to simulate at HEAD (XMILE-name + JSON ident boundaries skip name-context canonicalization, unlike the #690 serde fix) #702 -- freshly imported/created period-named variables may fail to simulate at HEAD
• engine: view-element name fields skip literal-period ident handling, so a period-named variable's diagram glyph/label/selection-association can mismatch (·-vs-.-vs-‥) #704 -- view-element name fields skip literal-period ident handling (glyph/label mismatch)
• engine: MDL writer round-trip is not idempotent (time_units gain, TIME BASE rewrite, NaN emission, wrong sim_specs model) #848 -- MDL writer round-trip is not idempotent (time_units gain, TIME BASE rewrite, NaN emission)
• engine: MDL roundtrip harness coverage gaps let confirmed writer bugs ship undetected #851 -- MDL roundtrip harness coverage gaps let confirmed writer bugs ship undetected
• engine: MDL importer subscript/dimension classification is non-deterministic across process hash seeds #859 -- MDL importer subscript/dimension classification is non-deterministic across hash seeds
• engine: MDL writer dim_at can silently widen a subrange reduction on misclassified wildcard #860 -- MDL writer dim_at can silently widen a subrange reduction on a misclassified wildcard
• engine: MDL export emits per-element graphical-function warnings one-per-element (should aggregate per variable) #861 -- MDL export emits per-element graphical-function warnings one-per-element
• engine: MDL writer units field re-import hard-fails on non-unit-expression text #862 -- MDL writer units field re-import hard-fails on non-unit-expression text
• to_mdl: initialized DELAY1/DELAY3 emitted without the I suffix (invalid Vensim) #886 -- to_mdl: initialized DELAY1/DELAY3 emitted without the I suffix (invalid Vensim)
• engine: MDL parser's binary operator precedence is inverted vs Vensim, masked by xmile_compat's precedence-free emission #914 -- MDL parser binary operator precedence is inverted vs Vensim (masked by xmile_compat)

Engine -- simulation builtins & numeric semantics

• Warn if constants don’t have units specified #62 -- warn if constants don't have units specified
• support modifying variables at each DT/runtime #71 -- support modifying variables at each DT / runtime
• engine: DELAY FIXED ring-buffer semantics blocks 4 simulation tests #346 -- DELAY FIXED ring-buffer semantics (blocks 4 sim tests)
• Synthesized _max aux can shadow user-authored variables #438 -- synthesized _max aux can shadow user-authored variables
• engine: resource_unit/resource_units alias not honored during unit inference #454 -- resource_unit/resource_units alias not honored during unit inference
• engine: unit checker cannot unify 'year' with the implicit 'time' unit of stocks #455 -- unit checker can't unify year with the implicit time unit of stocks
• engine: duplicate_unit diagnostics emitted once per stdlib module instance #456 -- duplicate_unit diagnostics emitted once per stdlib module instance
• engine: stdlib module output units not propagated to consumers #457 -- stdlib module output units not propagated to consumers
• engine: non_negative variable flag is parsed and stored but never enforced in the compiler or VM #545 -- non_negative variable flag is parsed/stored but never enforced
• engine: stochastic builtins (RANDOM NORMAL / NORMAL / UNIFORM / etc.) are recognized but unimplemented -- UnknownBuiltin at simulation, blocks ~5 metasd corpus models #560 -- stochastic builtins (RANDOM NORMAL / UNIFORM / ...) unimplemented
• engine: DELAY N / SMOOTH N require a compile-time-constant order -- non-constant/parameterized order -> UnknownBuiltin (blocks full simulation of thyroid-2008-d and any dynamic-order delay/smooth) #562 -- DELAY N / SMOOTH N require a compile-time-constant order
• engine: dormant vector.dat 2-D VECTOR SORT ORDER p block is inconsistent + genuine-Vensim multi-dim VSO semantics unverified by any live fixture #576 -- 2-D VECTOR SORT ORDER semantics unverified by any live fixture
• engine: VECTOR ELM MAP with scalar source + arithmetic (expression) offset fails to compile (no diagnostic) -- narrows the Phase 5 vector.xmile gate #578 -- VECTOR ELM MAP with scalar source + expression offset fails silently
• engine: VECTOR ELM MAP full_source_len has no genuine-Vensim numeric end-to-end test gate (defense-in-depth coverage gap) #579 -- VECTOR ELM MAP full_source_len has no numeric end-to-end test gate
• engine: INITIAL()-backed module/macro primary output omitted from the initials runlist (spurious inf) #584 -- INITIAL()-backed module/macro primary output omitted from initials runlist
• engine: arrayed VECTOR SORT ORDER returns flat global indices instead of per-iterated-slice 0-based ranks #585 -- arrayed VECTOR SORT ORDER returns flat global indices, not per-slice ranks
• engine: Vensim :NA: was mishandled as IEEE NaN (should be the finite sentinel -2^109) #586 -- Vensim :NA: mishandled as IEEE NaN (should be the finite sentinel)
• engine: per-element graphical functions are mapped by Arrayed-Vec (sorted) position instead of element-name -> dimension index #589 -- per-element graphical functions mapped by sorted position, not element name
• engine: C-LEARN data/graph-lookup variables import as 0+0 (zeroed) instead of their lookup/data values #590 -- C-LEARN data/graph-lookup variables import as 0+0 (zeroed)
• WASM/libsimlin: data-supply API for GET DIRECT external data (browser/WASM callers) #598 -- WASM/libsimlin data-supply API for GET DIRECT external data
• engine: remove legacy "0+0" LOOKUP_SENTINEL read-shim from lookup-only detection (gated on datamodel migration) #608 -- remove the legacy "0+0" LOOKUP_SENTINEL read-shim
• engine: INT/MOD negative-number semantics diverge from Vensim (blocks tests/rounding); MOD uses rem_euclid not XMILE floored modulus #610 -- INT/MOD negative-number semantics diverge from Vensim (floored modulus)
• engine: residual ~14 C-LEARN unit diagnostics need a Vensim-parity strictness decision (IF-branch / log-argument / coefficient cancellation) #613 -- ~14 residual C-LEARN unit diagnostics need a Vensim-parity strictness decision
• engine: variable named 'inf' (or other zero-arity builtin name) collides with the Inf builtin and silently reads +infinity #681 -- variable named inf collides with the Inf builtin, silently reads +infinity
• engine: Arrayed/ApplyToAll equation on a variable with no declared dimensions is silently collapsed to scalar instead of rejected #697 -- Arrayed/ApplyToAll with no declared dims silently collapses to scalar
• engine: graphical-function tables with unsorted/NaN x-coordinates are accepted silently and produce undefined lookup results #715 -- graphical-function tables with unsorted/NaN x-coords accepted silently
• engine: different-cardinality element-map reference fails to compile (failed to compile fragments) #753 -- different-cardinality element-map reference fails to compile ("failed to compile fragments")
• engine: executed A2A lowering ignores explicit element maps and resolves positionally (silent wrong results for permutation maps) #756 -- executed A2A lowering ignores explicit element maps and resolves positionally (silent wrong results)
• engine: mixed-arity multi-source reducer fails fragment compile (SUM(m1[d1,*]) + SUM(m2[*]) with 2-D m2) #781 -- mixed-arity multi-source reducer fails fragment compile (SUM(m1[d1,*]) + SUM(m2[*]))
• engine: per-source fan-out of unit inference diagnostics (enables per-field underlines deferred from #836) #837 -- per-source fan-out of unit-inference diagnostics (enables the per-field underlines deferred from diagram: unit inference errors underlined in the wrong place (UnitError kind collapsed to a boolean) #836)

Conveyors & queues (Stella conveyor/queue engine)

• Conveyor engine: deferred edge-case limitations (GF leak fraction, PREVIOUS(driven_flow), JSON serialization) #865 -- deferred edge-case limitations (GF leak fraction, PREVIOUS(driven_flow), JSON serialization)
• Queue engine: secondary/overflow outflow into a second conveyor bypasses that conveyor's batch/capacity rules #868 -- queue: secondary/overflow outflow into a second conveyor bypasses that conveyor's batch/capacity rules
• conveyor-engine: per-element <len>/conveyor attributes in <element> blocks have no representation, contra conveyors.md section 10 #904 -- per-element <len>/conveyor attributes in <element> blocks have no representation (conveyors.md 10)
• conveyor-engine: a mid-pass ConveyorTransitTooLong leaves belts 0..i already phase_a-advanced; run_to restores nothing #938 -- a mid-pass ConveyorTransitTooLong leaves belts 0..i already phase_a-advanced; run_to restores nothing
• engine: conveyor/queue in a sub-model is rejected, but the limitation is undocumented and both error paths cite a phantom conveyors.md 9.3 claim #940 -- conveyor/queue in a sub-model is rejected, undocumented, and both error paths cite a phantom conveyors.md 9.3
• test: sir_social_distancing_mixnot.stmx is unsimulatable by any backend (conveyor lives in a sub-model); #924's corpus gate would read the reject as expected #941 -- sir_social_distancing_mixnot.stmx is unsimulatable by any backend; wasm: drop the conveyor Unsupported reject and put conveyor fixtures under the parity harness #924's corpus gate reads the reject as expected
• engine: <leak_integers/> is silently ignored on an exponential-leak conveyor #942 -- <leak_integers/> is silently ignored on an exponential-leak conveyor
• xmile: conveyor ignore_earlier_zone_losses is unreachable -- the reader hardcodes false, and nothing writes it #943 -- ignore_earlier_zone_losses is unreachable -- the reader hardcodes false and nothing writes it
• engine: a direct read of a conveyor-driven leak flow is rejected, blocking a real Stella model (covid19_severity.stmx) #944 -- a direct read of a conveyor-driven leak flow is rejected, blocking covid19_severity.stmx
• wasmgen: conveyor inflow spreadflow placements (even/dest/dist/source) are rejected, not lowered #946 -- wasmgen: conveyor inflow spreadflow placements (even/dest/dist/source) are rejected, not lowered
• conveyor-engine: an integer leak flow into an ARRESTED destination belt can still deliver whole units #947 -- an integer leak flow into an ARRESTED destination belt can still deliver whole units
• engine: conveyor container slat access (belt[1]) simulates correctly but emits a spurious Error diagnostic #948 -- container slat access (belt[1]) simulates correctly but emits a spurious Error diagnostic
• test: <arrest> and <sample> have no conveyor corpus fixture (proto/XMILE round-trip untested) #949 -- <arrest> and <sample> have no corpus fixture (proto/XMILE round-trip untested)
• conveyor-engine: a queue-coupled discrete conveyor's slat contents are not integral, contra conveyors.md 6.4 rule 1 #950 -- a queue-coupled discrete conveyor's slat contents are not integral (conveyors.md 6.4 rule 1)
• test: the conveyor corpus has no independent numeric oracle -- a lockstep VM+wasm drift would pass every test #951 -- the corpus has no independent numeric oracle -- a lockstep VM+wasm drift would pass every test
• test: no corpus fixture for the belt apportion loop or for FIFO backlog / the four batch rules (README says #949 covers them; it does not) #954 -- no corpus fixture for the belt apportion loop or for FIFO backlog / the four batch rules

Engine -- incremental compilation, salsa, cycles & architecture

• further think about cycles in models #288 -- further think about cycles in models
• sync_from_datamodel creates new salsa inputs on every call, defeating cache reuse #290 -- sync_from_datamodel creates new salsa inputs every call, defeating cache reuse
• Thread persistent salsa DB through Project compilation for incremental patch validation #292 -- thread a persistent salsa DB through Project compilation
• Route compilation dependency analysis through salsa tracked queries #294 -- route compilation dependency analysis through salsa tracked queries
• Incremental compilation gaps: modules, builtins, variable layout #295 -- incremental compilation gaps (modules, builtins, variable layout)
• Incremental compiler panics on C-LEARN model (catch_unwind workaround) #363 -- incremental compiler panics on C-LEARN (catch_unwind workaround)
• engine: decouple module-aware parse context from per-variable parse cache #372 -- decouple module-aware parse context from the per-variable parse cache
• engine: protobuf ProjectPatch/ModelPatch/ModelOperation messages are dead schema #538 -- protobuf ProjectPatch/ModelPatch/ModelOperation messages are dead schema
• engine: AiInformation is silently dropped on every protobuf round-trip #539 -- AiInformation is silently dropped on every protobuf round-trip
• engine: SourceVariable.model_name is a misnamed correctness trap (it is the Module reference target, empty for non-Module vars -- not the owning model) #558 -- SourceVariable.model_name is a misnamed correctness trap
• engine: model.rs init/module dep split uses ASCII '.' but canonical module separator is U+00B7 (dead/incorrect branch) #567 -- model.rs init/module dep split uses ASCII . not the canonical U+00B7
• engine: two divergent circular-dependency cycle-gate code paths (System A salsa vs System B model.rs:442) #568 -- two divergent circular-dependency cycle-gate code paths
• engine: element-cycle dt-recurrence SCC resolution uses no-input build_var_info wiring; Phase 2 must plumb real module_input_names before consuming ResolvedScc.element_order #573 -- element-cycle dt-recurrence SCC uses no-input build_var_info wiring
• engine: no init-phase cycle instrumentation + engine-consistency cross-check (init analogue of dt_cycle_sccs / dt_cycle_sccs_consistency_violation) #574 -- no init-phase cycle instrumentation / engine-consistency cross-check
• engine: phase_element_order builds the SCC element graph from per-variable mini-slots -- zero cross-member edges + unsound resolution of genuine multi-variable element cycles (masked by members.len()!=1; Phase 2 Subcomponent B must remove mask) #575 -- phase_element_order SCC graph has zero cross-member edges (unsound)
• C-LEARN: post-cycle-gate assemble_simulation fragment-compile failure on synthetic temp-arg helpers #580 -- C-LEARN post-cycle-gate assemble_simulation fragment-compile failure
• engine: assembly-stage CompilationDiagnostics are silently discarded (IN_TRACKED_CONTEXT never set true) #581 -- assembly-stage CompilationDiagnostics silently discarded
• engine: GraphicalFunctionId u8 overflow in concatenate_fragments (no cross-fragment GF de-duplication) #582 -- GraphicalFunctionId u8 overflow in concatenate_fragments
• engine: TempId u8 overflow in incremental fragment assembly on large models (C-LEARN) #583 -- TempId u8 overflow in incremental fragment assembly (large models)
• engine: init runlist ordering is nondeterministic for an init-time algebraic cycle through a synthetic-module stock output #595 -- init runlist ordering nondeterministic for an init-time algebraic cycle
• engine: VM set_value cannot apply distinct per-instance overrides to two instances of the same shared CompiledModule #616 -- VM set_value can't apply per-instance overrides to a shared CompiledModule
• engine: two offset machineries disagree on submodel size with a table-only variable (latent off-by-N) #644 -- two offset machineries disagree on submodel size (latent off-by-N)
• engine: lower_implicit_var drops helper-fragment parse errors silently #666 -- lower_implicit_var drops helper-fragment parse errors silently
• engine: unguarded panics on dangling module model_name in legacy from_salsa path (module_deps, topo_sort) #806 -- unguarded panics on a dangling module model_name in the legacy from_salsa path
• engine: model_ltm_variables warnings duplicated across a module boundary in project error report #866 -- model_ltm_variables warnings duplicated across a module boundary in the project error report
• engine: per-model sim_specs are silently lost in protobuf persistence #934 -- per-model sim_specs are silently lost in protobuf persistence
• engine: StockFlow sketch compatibility is lost in protobuf persistence #936 -- StockFlow sketch compatibility is lost in protobuf persistence
• engine: Vm::into_results() reports slab capacity as step_count, not rows actually saved #939 -- Vm::into_results() reports slab capacity as step_count, not rows actually saved
• engine: model-level diagnostics leak the internal ModelError{code: details} Display form into user-facing messages #952 -- model-level diagnostics leak the internal ModelError{code: details} Display form to users

Engine -- performance

• Intern Ident<Canonical> strings via salsa::interned to reduce hashing and cloning overhead #317 -- intern Ident<Canonical> via salsa::interned to cut hashing/cloning
• Make UnitMap cheaply cloneable via Arc to reduce per-context allocation overhead #318 -- make UnitMap cheaply cloneable via Arc
• engine: threaded VM dispatch (guaranteed tail calls) to cut the dispatch branch-miss bottleneck #601 -- threaded VM dispatch (guaranteed tail calls) to cut branch-miss cost
• engine: optimize the graphical-function Lookup hot path (~4% of the C-LEARN run) #602 -- optimize the graphical-function Lookup hot path (~4% of C-LEARN run)
• engine: stride-step RuntimeView::flat_offset for sequential iteration #603 -- stride-step RuntimeView::flat_offset for sequential iteration
• engine: bundle marginal branch/dispatch reductions to test for a CPU predictor threshold #604 -- bundle marginal branch/dispatch reductions to test a predictor threshold
• engine: salsa::Update on compiled bytecode uses PartialEq over f64/NaN (benign missed-backdate) #642 -- salsa::Update on bytecode uses PartialEq over f64/NaN (benign missed-backdate)
• engine: compile_project_incremental deep-clones CompiledSimulation out of the Arc on every call #643 -- compile_project_incremental deep-clones CompiledSimulation each call
• engine: lazy IF evaluation — skip the untaken branch in the VM #711 -- lazy IF evaluation -- skip the untaken branch in the VM
• engine: time-invariant variable hoisting — a constant phase evaluated once per run_to #712 -- time-invariant variable hoisting -- constant phase evaluated once per run

Engine -- hygiene, dead code, lint caps & wasm-backend debt

• engine: audit and clean up #[allow(dead_code)] annotations #424 -- audit and clean up #[allow(dead_code)] annotations
• engine: vdf.rs exceeds MAX_LINES=6000 lint threshold (6001 lines), breaking pnpm lint #471 -- vdf.rs exceeds MAX_LINES (6001) breaking pnpm lint
• tools/vdf_xray.py stubs dataset (0x41) parsing while Rust VdfDatasetFile is implemented #501 -- tools/vdf_xray.py stubs dataset parsing while Rust VdfDatasetFile exists
• engine: db.rs is 1 line from the 6000-line lint cap; split before it blocks an unrelated commit #518 -- db.rs is 1 line from the 6000-line cap; split it preemptively
• engine: macro-attributable Diagnostic classifier is duplicated across a crate compilation boundary (lockstep enforced only by a comment) #563 -- macro-attributable Diagnostic classifier duplicated across a crate boundary
• engine: three visually-indistinguishable separator chars (U+205A / U+00B7 / ASCII .) are load-bearing across stdlib + macro subsystems with no type-level or lint guard #565 -- three indistinguishable separator chars load-bearing with no guard
• wasm-backend plan: false 'CompiledSimulation is fully un-fused' premise misleads un-executed phases 2-8 #607 -- wasm-backend plan: false "CompiledSimulation is fully un-fused" premise
• wasm-backend Phase 5: broadcast-iteration opcode family is unreached dead code; ViewRangeDynamic unsupported by design (neither in corpus) #612 -- wasm-backend Phase 5 broadcast-iteration opcode family is dead code
• wasm-backend AC3.2 boundary: inline-fixture VM tests bypass the wasm parity hook #617 -- wasm-backend AC3.2: inline-fixture VM tests bypass the wasm parity hook
• ts: two divergent canonicalize implementations (@simlin/core incomplete vs @simlin/engine Rust-faithful) #624 -- two divergent TS canonicalize implementations (@simlin/core vs /engine)
• engine: LTM-off Sim.getRun() now returns Run.links=[] instead of populated-scoreless (AC1.3 deviation) #626 -- LTM-off Sim.getRun() returns Run.links=[] (AC1.3 deviation)
• engine: VM vs wasm diverge on getValue after run_to(target) past FINAL_TIME #634 -- VM vs wasm diverge on getValue after run_to past FINAL_TIME

Layout engine

• Layout engine drops module view elements during diagram regeneration #340 -- layout drops module view elements during diagram regeneration
• layout: classify_flow_sides assumes left-to-right chain direction #445 -- classify_flow_sides assumes left-to-right chain direction
• layout: reorder_attachments_by_position loses visual ordering when flows change faces #446 -- reorder_attachments_by_position loses ordering when flows change faces
• layout: revisit AC1.8 scale-invariance term classification during Phase 4 calibration #631 -- revisit AC1.8 scale-invariance term classification during calibration
• layout: generate_layout_with_config is nondeterministic per seed (HashMap/HashSet iteration order in force accumulation) #633 -- generate_layout_with_config is nondeterministic per seed (HashMap order)
• layout: calibrate flow-valve link crossing metric #638 -- calibrate the flow-valve link-crossing metric
• layout: dependency-graph extraction misses lookup-table references, so lookup tables appear as isolated nodes with no connectors #650 -- dependency-graph extraction misses lookup-table refs (isolated nodes)

Frontend / diagram / editor

• Support creating aliases #43 -- support creating aliases
• snap to alignment #47 -- snap to alignment
• multi-part flows #51 -- multi-part flows
• if diagram is completely(ish) offscreen on current viewport (on didMount) center it #52 -- center the diagram on didMount when it's (mostly) offscreen
• when moving a cloud, the flow label jumps around #53 -- flow label jumps around when moving a cloud
• show error when diagram is out of sync with equations #54 -- show an error when the diagram is out of sync with equations
• every character change in sim specs shouldn't result in a model save #55 -- every character change in sim specs shouldn't trigger a model save
• multi-part connectors #60 -- multi-part connectors
• 'Cancel' not prevent equation from being edited #86 -- 'Cancel' doesn't prevent an equation from being edited
• Flows are invisible on creation, if created left or up #110 -- flows are invisible on creation if created left or up
• diagram: cleanup Point2D #229 -- diagram: cleanup Point2D
• diagram: selection-mutating Editor handlers bypass handleSelection, skipping showDetails reset and onSelectionChanged #529 -- selection-mutating Editor handlers bypass handleSelection
• Canvas view changes trigger a full engine round-trip per frame during pan/zoom #707 -- canvas pan/zoom triggers a full engine round-trip per frame
• rsbuild/rspack tree-shakes reset.css side-effect import out of src/app production CSS bundle #708 -- rsbuild tree-shakes reset.css out of the production CSS bundle
• diagram: component-library CSS modules hardcode light colors with no dark-mode overrides #709 -- component-library CSS hardcodes light colors with no dark-mode overrides
• core: view-element fidelity fields dropped on view edits (ViewElementCompat, View.font, group MDL marker, View.name) #811 -- core: view-element fidelity fields dropped on view edits (ViewElementCompat, View.font, group MDL marker, View.name)
• diagram: stale pre-edit sim curves shown (no staleness indicator) after an edit makes the model non-simulatable #813 -- stale pre-edit sim curves shown, with no staleness indicator, once an edit makes the model non-simulatable
• diagram: units editor accepts Enter, inserting literal newlines into the units string #829 -- units editor accepts Enter, inserting literal newlines into the units string
• Flow endpoint detach off a stock has no live preview (only valve moves) #830 -- flow endpoint detach off a stock has no live preview (only the valve moves)
• diagram: adjustFlows valve-mirroring is non-idempotent for off-center valves on the min-coordinate endpoint #832 -- source-cloud grab reflects an off-center flow valve on the first drag frame
• diagram: applySimSpecChange full-payload read-modify-write can silently revert a concurrent sim-specs commit (lost update) #833 -- applySimSpecChange full-payload read-modify-write can lose a concurrent sim-specs commit
• diagram: Button silently drops unknown props (dead data-testid in ModuleDetails) #959 -- Button silently drops unknown props (dead data-testid in ModuleDetails)
• app: NewUser marks the Username field invalid for the unrelated terms-agreement error #961 -- app: NewUser marks the Username field invalid for the unrelated terms-agreement error
• diagram: sim-specs fields silently revert invalid input with no feedback #962 -- sim-specs fields silently revert invalid input with no feedback

Server / backend / auth & security

• run management #41 -- run management
• rename models #48 -- rename models
• Add COOP and COEP headers #63 -- add COOP and COEP headers
• Cannot log in locally #78 -- cannot log in locally (local OAuth setup; undocumented Linux/WSL prerequisites)
• Real-time collaboration #106 -- real-time collaboration
• Full app local dev still requires Firebase (no lightweight/SQLite backend) #109 -- needs heavy libraries + a Google account to run locally (triage: partially addressed by @simlin/serve)
• /auth/providers endpoint enables email enumeration; add rate limiting #380 -- /auth/providers enables email enumeration; add rate limiting
• Move hardcoded Firebase API key to environment configuration #381 -- move the hardcoded Firebase API key to environment configuration
• Remove deprecated FirestoreAuthStrategy and POST /session endpoint #382 -- remove the deprecated FirestoreAuthStrategy and POST /session endpoint
• Clean up CSP frame-src directive after server-side auth migration #383 -- clean up the CSP frame-src directive after the auth migration
• OAuth login buttons render even when provider credentials are not configured #384 -- OAuth login buttons render even without provider credentials
• Apple OAuth flow should include nonce to prevent ID token replay #385 -- Apple OAuth flow should include a nonce to prevent ID-token replay
• server: ensure Firestore composite index for (providerUserId, provider) on users collection #386 -- ensure a Firestore composite index for (providerUserId, provider)
• Watcher validation gate is a no-op for never-opened entries (first-touch baseline bug) #478 -- watcher validation gate is a no-op for never-opened entries
• server: hosted-project saves leak orphaned File docs in Firestore (no version chain, no GC) #521 -- hosted-project saves leak orphaned File docs in Firestore (no GC)
• Logout flow is unimplemented (client menu item is a no-op; DELETE /session is a non-responding stub) #524 -- logout flow is unimplemented (menu no-op; DELETE /session a stub)
• embed: cross-origin web component broken -- module worker chunk fetched from app.simlin.com with no CORS/fallback #688 -- embed: cross-origin web component broken (no CORS/fallback)
• server: @google-cloud/trace-agent .start() is unguarded at boot and the package is a maintenance-mode dead end #689 -- @google-cloud/trace-agent .start() unguarded at boot; dead-end package
• Login email flow uses deprecated fetchSignInMethodsForEmail; Email Enumeration Protection toggle locks out all email/password users #692 -- login email flow uses deprecated API; enumeration toggle locks users out
• ops: add minimum production monitoring (notification channel, healthz route + uptime checks, log-based ERROR alerts, 5xx/latency alerts) #693 -- add minimum production monitoring (healthz, uptime, error/5xx alerts)
• server: preview rendering lost worker-thread isolation; no timeout + unbounded GAE instances compound the DoS/cost surface #694 -- preview rendering lost worker-thread isolation (DoS/cost surface)
• server: no SIGTERM/graceful-shutdown handler drops in-flight saves on GAE recycle #814 -- no SIGTERM / graceful-shutdown handler; in-flight saves are dropped on a GAE recycle
• serve: external edits made during watcher startup are silently and permanently lost #918 -- simlin-serve: external edits made during watcher startup are silently and permanently lost
• server: production dependency audit flags vulnerable uuid 9.0.1 #937 -- production dependency audit flags vulnerable uuid 9.0.1

Build / lint / type-check tooling

• engine: tsbuildinfo cache prevents incremental rebuild after clean #285 -- tsbuildinfo cache prevents incremental rebuild after clean
• core: tsconfig.json excludes test files, breaking eslint #441 -- core tsconfig.json excludes test files, breaking eslint
• codex-cli 0.121.0 fails to parse user-invokable key in SKILL.md frontmatter, blocking codex-review #462 -- codex-cli 0.121.0 can't parse user-invokable in SKILL.md frontmatter
• build: factor shared cross-build infrastructure between simlin-mcp and simlin-serve into scripts/lib/ #473 -- factor shared cross-build infra between simlin-mcp and simlin-serve
• docs/sdai-model.schema.json drifts from generator output (no CI gate) #475 -- sdai-model.schema.json drifts from generator output (no CI gate)
• build/pysimlin: CFFI extension rebuild silently reuses a stale libsimlin.a (no static-lib dependency tracking) #682 -- CFFI extension rebuild silently reuses a stale libsimlin.a
• build: React packages have no rules-of-hooks / exhaustive-deps linting #718 -- React packages have no rules-of-hooks / exhaustive-deps linting
• check-docs.py validates path references against the filesystem, so local pre-commit and clean-checkout CI disagree #725 -- check-docs.py makes local pre-commit and clean-checkout CI disagree
• examples/svg.mjs and scripts/debug-diagram-gen.mjs fail with ERR_MODULE_NOT_FOUND: @simlin/engine #827 -- examples/svg.mjs and scripts/debug-diagram-gen.mjs fail with ERR_MODULE_NOT_FOUND
• build: no dist-level smoke check that @simlin/diagram CSS (theme.css/reset.css) lands in serve-web bundle #828 -- no dist-level smoke check that @simlin/diagram CSS lands in the serve-web bundle
• build: eslint builds a full TypeScript Program (parserOptions.project) but enables zero type-aware rules #892 -- eslint builds a full TypeScript Program (parserOptions.project) but enables zero type-aware rules
• build: no-undef is enabled on TypeScript files, silently masked by parserOptions.project #893 -- no-undef is enabled on TypeScript files, silently masked by parserOptions.project
• build: enable typescript-eslint promise-safety rules -- ~20 real defects are invisible today #894 -- enable typescript-eslint promise-safety rules -- ~20 real defects are invisible today
• build: eslint.config.shared.js mutates module-level shared objects (cross-call clobbering, fragile index lookup) #895 -- eslint.config.shared.js mutates module-level shared objects (cross-call clobbering)
• build: eslint.config.shared.js requires five packages the root package.json never declares (resolves only via pnpm's virtual store) #896 -- eslint.config.shared.js requires five packages the root package.json never declares
• build: TypeScript test files are never type-checked (~55 latent type errors) #899 -- TypeScript test files are never type-checked (~55 latent type errors)
• serve: simlin-serve/web builds with Vite while the rest of the repo builds with Rsbuild #901 -- simlin-serve/web builds with Vite while the rest of the repo builds with Rsbuild
• build: rstest 'prefer-alias' disables tsconfig paths, silently resolving unaliased @simlin/* imports to stale built lib/ #911 -- rstest prefer-alias disables tsconfig paths, resolving @simlin/* to stale built lib/
• build: 108 rustdoc warnings workspace-wide with no CI gate on cargo doc #917 -- 108 rustdoc warnings workspace-wide with no CI gate on cargo doc
• build: cargo-deny is blocked by RUSTSEC-2026-0204 in crossbeam-epoch #932 -- cargo-deny is blocked by RUSTSEC-2026-0204 in crossbeam-epoch
• build: rstest node environment silently breaks named imports from goog-style CJS modules (google-protobuf *_pb) #956 -- rstest node environment silently breaks named imports from goog-style CJS modules (*_pb)
• build: per-package needs-format/format prettier scripts glob generated lib/ output (server, diagram); server also errors on a non-matching tsx pattern #957 -- per-package prettier scripts glob generated lib/ output; server errors on a non-matching pattern

CI / test hygiene

• engine: wrld3_resource_unit_alias_should_not_conflict is order-dependent / flaky #474 -- wrld3_resource_unit_alias_should_not_conflict is order-dependent / flaky
• server: @simlin/server Jest suite intermittently fails one test under parallel pre-commit load #635 -- @simlin/server Jest suite intermittently fails one test under load
• engine: test_render_arc_svg_byte_identical fails on main from ~1 ULP float-formatting drift; byte-identical SVG golden test is fragile #639 -- byte-identical SVG golden test fragile (~1 ULP float-formatting drift)
• ci: add a release-mode lane so C-LEARN-scale tests can be un-ignored #657 -- add a release-mode CI lane so C-LEARN-scale tests can be un-ignored
• test: no automated cross-origin embed regression test for sd-component.js (follow-up to #688) #825 -- no automated cross-origin embed regression test for sd-component.js (follow-up to embed: cross-origin web component broken -- module worker chunk fetched from app.simlin.com with no CORS/fallback #688)
• test: nothing runs the simlin-engine #[ignore]d tier; 7 tests fail there today (one untracked) #915 -- nothing runs the simlin-engine #[ignore]d tier; 7 tests fail there today
• test: simlin-serve watcher integration tests are flaky on macOS (fixed wall-clock timeouts vs FSEvents latency) #916 -- simlin-serve watcher integration tests are flaky on macOS (fixed timeouts vs FSEvents latency)
• test: main-app Playwright suite is stale, unrunnable, and absent from CI #926 -- main-app Playwright suite is stale, unrunnable, and absent from CI
• test: simlin-serve watcher test runtime scales with fseventsd health, can bust the pre-commit cap on passing tests #945 -- simlin-serve watcher test runtime scales with fseventsd health; can bust the pre-commit cap

Deploy / ops / release / CLI

• FR: Simlin-engine as separate crate #91 -- publish simlin-engine to crates.io as a standalone crate (external request)
• AddressSanitizer: resvg 0.47.0 type mismatch on nightly Rust #287 -- AddressSanitizer: resvg 0.47.0 type mismatch on nightly Rust
• cli debug and print_tsv_comparison silently truncate on length mismatch #437 -- cli debug / print_tsv_comparison silently truncate on length mismatch
• simlin-cli: add --rounds flag for systems format models #439 -- simlin-cli: add a --rounds flag for systems format models
• pysimlin: setuptools_scm fails on mcp-v0.1.4 git tag #442 -- pysimlin: setuptools_scm fails on the mcp-v0.1.4 git tag
• deploy: .gcloudignore doesn't import .gitignore; gcloud app deploy uploads untracked files and trips GAE 10k-file cap #695 -- .gcloudignore doesn't import .gitignore; trips the GAE 10k-file cap
• deploy: static-handler CSP/HSTS headers can silently drift out of .app.prod.yaml (and validator skips them + runtime) #815 -- static-handler CSP/HSTS headers can drift out of .app.prod.yaml (validator and runtime both skip them)
• Harden the deploy-assembly CI gate: Node-major skew, unpinned engines.node, non-frozen install, staged path untested #816 -- harden the deploy-assembly CI gate: Node-major skew, unpinned engines.node, non-frozen install, staged path untested
• cli: convert -o output write is non-atomic and panics on IO errors #855 -- cli: convert -o output write is non-atomic and panics on IO errors
• cli: stdout/stderr stream discipline is documented but unpinned by any test #953 -- cli: stdout/stderr stream discipline is documented but unpinned by any test

MCP / AI

• In MCP responses we need to sample time series when they're too long #337 -- sample time series in MCP responses when they're too long
• SD-AI relationships may be incomplete for models with pre-existing equation errors #433 -- SD-AI relationships may be incomplete for models with equation errors

Docs

• Need documentation on how to import a Vensim mdl model. #102 -- documentation on how to import a Vensim mdl model
• Prune ephemeral implementation plans from docs/design-plans/ #407 -- prune ephemeral implementation plans from docs/design-plans/
• docs: implementation-plans/2026-05-18-element-cycle-resolution not in docs/README.md index (docs/CLAUDE.md convention violation) #577 -- add the element-cycle-resolution plan to the docs/README.md index
• docs: reconcile the two parallel implementation-plan locations (docs/plans vs docs/implementation-plans) and fix stale path references left by the plan-removal cleanup #593 -- reconcile the two implementation-plan locations + fix stale path refs
• docs: src/simlin-cli/CLAUDE.md has a banned "Last updated" line #920 -- src/simlin-cli/CLAUDE.md has a banned "Last updated" line

---

Deferred to dedicated trackers

• LTM (35 open): epic: Loops that Matter (LTM) -- tracking known issues and improvements #488 -- gh issue list --label ltm --state open
• Array support: array support #25 | Unit checking: unit checking #35 | Module editing: module editing #46 | Fork & merge: fork and merge models #57

Conventions

• New bug/improvement: file a normal issue with area label(s) and add a - [ ] #N -- title row to the matching section (or to epic: Loops that Matter (LTM) -- tracking known issues and improvements #488 if LTM).
• When closing a sub-issue, tick its box here in the same PR so the human view stays in sync.
• Quick-win hunting: gh issue list --label hygiene --state open and gh issue list --label bug --state open are the most self-contained.