docs(quest): record TS duration reproduction evidence

[kixelated]

The previously reported TS export captured about 3.3 seconds from a 20.1-second input. An isolated current-main reproduction captured 20.47 seconds from a 20.06-second source (ratio 1.021), and the publisher exited successfully.

This draft records that evidence in the quest rather than inventing a duration fix. The harness instead failed a distinct T-STD hold-time check: PID 257 held data for more than one second. Existing export end-following work (#5147) and relay group-completion work (#5175) remain separate.

Validation: nix develop --command just test ts reproduced the T-STD failure, passed duration fidelity, and retained capture and publisher/receiver logs. quest check passed. The next useful step is to capture a duration-failing run after those existing fixes, including publisher exit status and relay logs. No thresholds, retries, or timeouts changed.

Public API and wire impact: none. The quest remains open; this is an investigation draft, not a completed bug fix.

(Written by GPT-6)

[kixelated]

Closing this draft. #5253 moved the quest to quest/m0/ts-duration-fidelity.md and folded in this draft's evidence: an isolated local run captured the whole source (20.47 s of 20.06 s), while CI Interop fails on every run, main included. The m0 quest now treats it as a regression and starts by bisecting from the last green Interop run.

(Written by Claude Opus 5.5)