Doc-only. Numbers re-measured on 6004398, not inherited — this file's own rule.
before
after
line
92.8% (2183/2352)
94.2% (2234/2372)
branch
81.3% (1091/1342)
87.5% (1171/1338)
tests
584 in 87 classes
628 in 96
missed
169 lines, 251 branches
138 lines, 167 branches
The entry explains its own branch move
Because this file has a documented history of not doing that. Wave 4's rise is almost entirely numerator — up 80 (1091 → 1171) with the denominator moving −4 (1342 → 1338). That is unlike the
2026-08-29 seam work, where the numerator rose 37 (974 → 1011) while the denominator fell 70
(1410 → 1340), so much of that gain was scaffolding leaving the measurement rather than arms being
covered. The line denominator grew the other way, 2352 → 2372, and that is the seams this wave cut — capabilitiesFrom, ffprobeInfoFrom, sessionOutcome, sweepScope/startupSweep — not new
untested code.
(An earlier draft of this attributed the 37/−70 decomposition to wave 3. It belongs to the #153
children. Corrected before commit.)
A methodological result, recorded because it outlives the percentages
#218 fixed an intermittent flake. Running the whole suite six times per arm caught nothing in either
arm — at the observed rate a clean six-run arm is roughly a coin flip, so it was underpowered and
proved nothing. A deterministic mutation settled it, and the merge train then confirmed it by
accident when the race reproduced on #217's Unit tests leg, which sits below the fix. Hence: prefer a mutation that must go red to a repetition count when the fix is for something
intermittent.
The watchdog trap gains its outcome
The JUnit-timeout entry has always said the jstack watchdog is the only attribution a hung run gets.
It has now fired: it caught #125's Room/WorkManager deadlock on CI in 10m57s with the hung test
named, against that ticket's own prediction of a 60-minute cap and no cause. #125 is closed as
bounded — the inversion is internal to room/work-runtime and still live at 2.11.2 / 2.7.0.
Not touched
Instrumented counts.#221 adds a test and so moves "three of the 60" to 61; that PR carries the
prose, so this file never claims a count main does not have.
Doc-only. Numbers **re-measured on `6004398`**, not inherited — this file's own rule.
| | before | after |
|---|---|---|
| line | 92.8% (2183/2352) | **94.2%** (2234/2372) |
| branch | 81.3% (1091/1342) | **87.5%** (1171/1338) |
| tests | 584 in 87 classes | **628 in 96** |
| missed | 169 lines, 251 branches | **138 lines, 167 branches** |
## The entry explains its own branch move
Because this file has a documented history of not doing that. Wave 4's rise is almost entirely
**numerator** — up 80 (1091 → 1171) with the denominator moving −4 (1342 → 1338). That is unlike the
2026-08-29 seam work, where the numerator rose 37 (974 → 1011) while the denominator *fell* 70
(1410 → 1340), so much of that gain was scaffolding leaving the measurement rather than arms being
covered. The line denominator grew the other way, 2352 → 2372, and that is the seams this wave cut —
`capabilitiesFrom`, `ffprobeInfoFrom`, `sessionOutcome`, `sweepScope`/`startupSweep` — not new
untested code.
*(An earlier draft of this attributed the 37/−70 decomposition to wave 3. It belongs to the #153
children. Corrected before commit.)*
## A methodological result, recorded because it outlives the percentages
#218 fixed an intermittent flake. Running the whole suite six times per arm caught nothing in either
arm — at the observed rate a clean six-run arm is roughly a coin flip, so it was underpowered and
proved nothing. A deterministic mutation settled it, and the merge train then confirmed it by
accident when the race reproduced on #217's Unit tests leg, which sits below the fix. Hence:
**prefer a mutation that must go red to a repetition count** when the fix is for something
intermittent.
## The watchdog trap gains its outcome
The JUnit-timeout entry has always said the jstack watchdog is the only attribution a hung run gets.
It has now fired: it caught #125's Room/WorkManager deadlock on CI in 10m57s with the hung test
named, against that ticket's own prediction of a 60-minute cap and no cause. #125 is closed as
bounded — the inversion is internal to `room`/`work-runtime` and still live at 2.11.2 / 2.7.0.
## Not touched
**Instrumented counts.** #221 adds a test and so moves "three of the 60" to 61; that PR carries the
prose, so this file never claims a count `main` does not have.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Doc-only. Numbers re-measured on
6004398, not inherited — this file's own rule.The entry explains its own branch move
Because this file has a documented history of not doing that. Wave 4's rise is almost entirely
numerator — up 80 (1091 → 1171) with the denominator moving −4 (1342 → 1338). That is unlike the
2026-08-29 seam work, where the numerator rose 37 (974 → 1011) while the denominator fell 70
(1410 → 1340), so much of that gain was scaffolding leaving the measurement rather than arms being
covered. The line denominator grew the other way, 2352 → 2372, and that is the seams this wave cut —
capabilitiesFrom,ffprobeInfoFrom,sessionOutcome,sweepScope/startupSweep— not newuntested code.
(An earlier draft of this attributed the 37/−70 decomposition to wave 3. It belongs to the #153
children. Corrected before commit.)
A methodological result, recorded because it outlives the percentages
#218 fixed an intermittent flake. Running the whole suite six times per arm caught nothing in either
arm — at the observed rate a clean six-run arm is roughly a coin flip, so it was underpowered and
proved nothing. A deterministic mutation settled it, and the merge train then confirmed it by
accident when the race reproduced on #217's Unit tests leg, which sits below the fix. Hence:
prefer a mutation that must go red to a repetition count when the fix is for something
intermittent.
The watchdog trap gains its outcome
The JUnit-timeout entry has always said the jstack watchdog is the only attribution a hung run gets.
It has now fired: it caught #125's Room/WorkManager deadlock on CI in 10m57s with the hung test
named, against that ticket's own prediction of a 60-minute cap and no cause. #125 is closed as
bounded — the inversion is internal to
room/work-runtimeand still live at 2.11.2 / 2.7.0.Not touched
Instrumented counts. #221 adds a test and so moves "three of the 60" to 61; that PR carries the
prose, so this file never claims a count
maindoes not have.🤖 Generated with Claude Code