feat(BACKEND-TENSTORRENT): tt_clock_state sampler/judge + busy-slice refold (#2005); #2003 re-adjudicated clock-attributed

[lu-zero]

record(BACKEND-TENSTORRENT): #2003 re-adjudicated clock-attributed — inversion stands at governor parity; tt_clock_state first wired use (#2005)

The tool landed same-day and its first consumer was the morning's
unattributed result. Same binary, workload, alternation and lock discipline;
one sampler window per arm around six order-alternated legs.

Throughput: default host-free eager median 10.880 tok/s vs opt-out 13.645
(ratio 1.254) — reproduces the morning's 10.822 vs 13.369.

Attribution finding: the Blackhole P150 AICLK governor is two-state on this
telemetry, 800 idle vs pegged under load, so all six raw windows refuse
within-run spread at 40.74% — kept verbatim in the evidence log, because
loosening the NVIDIA rule to fit one board is exactly the failure mode the
protocol forbids. Attribution instead rides the live-recorded busy column:
pid-held /dev/tenstorrent fds per interval, recorded before any outcome
value existed. tt_refold_busy.py rebuilds those slices and every window
shows ONE distinct AICLK {1350}, spread 0.00%, cross-arm offsets 0/0,
judge PASS rc=0. Both arms computed at identical cap clocks; the inverted
ratio is a real path difference and #2003 stays open for the per-op delta.

Also on this branch: the consolidated host-free gap row (supersedes either
single-session wording regardless of merge order against #2004), the #2005
index row naming its owed items — verified claimed-max pin, in-process
pyluwen sampling, and the spread-scope policy decision — and the spec's
Evidence/Owed sections carrying both findings and the explicit
never-loosen-a-threshold note.

Evidence verbatim: docs/bench-evidence/tt-p150-clock-attributed-20260826.log

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: AGENT:zai-glm-5.3-flash [maki]

[lu-zero]

Rebased onto origin/main 2a42cb369 (KV-GDN-STATE-BUDGET), clean. Head now b3182637b:

• 5 previous commits replayed unchanged in content
• new tail commit closes the three gaps the first fresh-review pass found: red-first tests pin refold busy-filter collapse (mutation (a)), ratio/offset arithmetic vs synthetic fixtures (mutation (c)), and tt_refold_busy.py is now reached by tests ((e)). Suite 16→22 cases; tools suite 539 green
• role-discipline gate initially reddened the repair commit because PR references are read from the SUBJECT only; re-authored with (feat(BACKEND-TENSTORRENT): tt_clock_state sampler/judge + busy-slice refold (#2005); #2003 re-adjudicated clock-attributed #2006)

Stacking note: this PR supersedes parts of #2004's morning-unattributed wording (host-free row, benchmark-record entries) with the clock-attributed re-adjudication. Mechanical consequence, verified by merge simulation in both directions in review: git conflicts on those record lines whichever lands first. Intended order: #2004 first (self-consistent morning snapshot), then this PR rebases over it taking the target rows and reapplying the scoped clock-attributed edits, per the keyed-record rule.