quest: plan the Interop latecomer control and TS PSI runs

quest/m0/README.md

@@ -61,6 +61,7 @@ remain and no release waits on them.
 - [CI host](/quest/m0/ci-host.md) - the maintainer brings up the spare desktop as the `moq-ci` and `moq-gpu` runner host
 - [Draft-22 media on 0.17](/quest/m0/release-22/README.md) - a 0.17.x with the LOCATION_FILTER and FIRST_OBJECT fixes and moq-noq 1.3.5, before Seattle
 - [TS duration fidelity](/quest/m0/ts-duration-fidelity.md) - Interop's TS compliance captures the whole round-tripped stream again, fixing a regression that turns Interop red on main
+- [Interop latecomer control](/quest/m0/interop-latecomer-control.md) - the lagging-latecomer control fails as designed, and a red Interop step no longer skips the TS steps
 - [Paused spinner](/quest/m0/watch-paused-spinner.md) - the watch buffering spinner never covers the paused play button, so Interop's resume step can click it
 - [Check base](/quest/m0/check-base.md) - `just check` diffs against the PR's base whatever the local branch tracks, so scoped checks stay scoped
 - [web-transport releases the qmux fixes](/quest/m0/qmux-credit-upstream.md) - waiting on moq-dev/web-transport#412 and #413 to merge and ship, which qmux credit bumps to

quest/m0/interop-latecomer-control.md

@@ -0,0 +1,31 @@
+# [S] Interop's lagging latecomer control fails as designed
+
+## Goal
+
+Interop's "control: lagging latecomer" negative control reliably fails
+"late join reaches live", so a passing positive case means something. A
+failing Interop step no longer hides the steps after it: the TS steps run
+and report on their own whatever the browser media step did.
+
+## Plan
+
+Found on #5207 (2026-10-10). The control (`test/interop/interop.sh`, added in
+#5173) runs `test/interop/clients/js/media.ts --lag --cases late-join
+--expect-fail "late join reaches live"`. In 3 of 4 runs on #5207, and on
+unrelated PRs, the lagged latecomer still reached live, so the control
+failed. Because the media step fails first, CI skips every TS Interop step,
+so TS regressions can land with no CI signal (the `--headroom` arm added in
+#5207 has not run in CI yet).
+
+- Find why `--lag` doesn't keep the player behind `LIVE_LAG_FRAMES` (the
+  player may catch up, the delay may not apply, or live detection may be too
+  loose), and fix that at its cause. Never widen thresholds, add retries, or
+  lengthen timeouts.
+- Decided (2026-10-10): in this quest, make the Interop steps independent, so
+  one red step never skips another. The job still fails if any step fails.
+
+Public API: none. Wire: none.
+
+## Related
+
+- [TS duration fidelity](/quest/m0/ts-duration-fidelity.md) - the other failure that keeps Interop red on main

quest/m1/tstd/README.md

@@ -37,6 +37,7 @@ nightly.
 
 - [Send-ahead within the delay](/quest/m1/tstd/send-ahead.md) - total lag is `--delay`, send-ahead included, with a 1 s default
 - [Mux-rate hold](/quest/m1/tstd/mux-rate-hold.md) - import publishes its catalog once the mux rate is measured, so export is constant-rate from the start
+- [PSI runs](/quest/m1/tstd/psi-runs.md) - aligned keyframes never lay identical PAT/PMT runs back to back, so the system buffer holds
 
 ## Related
 

quest/m1/tstd/psi-runs.md

@@ -0,0 +1,26 @@
+# [S] A TS slot never repeats a PSI run back to back
+
+## Goal
+
+When keyframes on several video PIDs align in one slot, the export does not
+lay identical PAT and PMT runs back to back, so the system buffer (512 bytes,
+ISO 13818-1 2.4.2.3) never overflows at 25 Mb/s and up. Each send-ahead unit
+still leads with its own tables as a tune-in point, and changed tables (a new
+PMT version) always go out, in order.
+
+## Plan
+
+Found by Codex on #5207 (`schedule.rs`, repeated PSI groups). It exists on
+`main` before #5207 too. Keeping each table once per slot was tried and broke
+`aac_program_config_follows_each_table`, because send-ahead units need their
+own tables.
+
+- Start from the author's narrower diff on #5207
+  (https://github.com/moq-dev/moq/pull/5207#discussion_r4239051134): drop a run
+  only when the same flush would lay it right after an identical run, and pad
+  the dropped packets back with nulls in a padded slot.
+- Test against system-buffer fill with the T-STD model, not a packet count:
+  three or more aligned video keyframes at 25 Mb/s overflow today, and pass
+  after the fix. The AAC tune-in test still passes.
+
+Public API: none. Wire: none.