Someone else's agent found the bug; I decided the fix

## Spark

From the 2026-08-22 digest. In one week on `zfs-replicate` I reviewed
outside pull requests proposing fixes for two defects, then landed my own
pull requests for the same two defects.

## Why it could be interesting

The accountability argument usually runs on hypotheticals. This is a
commit history where the work and the decision visibly separate: another
party's contribution identified the defect and offered a patch, and the
merge decision, the framing of the fix, and the resulting issues were
mine. That's the division the abstract argument keeps reaching for, in a
form that can be linked to rather than asserted.

## Open questions

- [ ] What actually differed between the outside patches and the ones
that landed — correctness, scope, house style, or nothing much?
- [ ] Does reviewing a contribution you could have written yourself cost
more or less than writing it?
- [ ] One of the outside PRs added a Vale style for instruction files —
the contribution was a sensor, not a fix. Does that change the
calculus?
- [ ] Can this be written without over-claiming what the contributors
were? Establish who and what before asserting anything.

## Source material

- — outside PR: preserve remote snapshot lookup key
- — the fix that landed: key remote snapshots by local filesystem
- — the issue behind it
- — outside PR: f-string in SNAPSHOTS filesystem strategy
- — the fix that landed: vary filesystem names in SNAPSHOTS
- and — the follow-on issues
- — outside PR adding a Vale style for instruction files
- Adjacent open ideas: #357 (can an agent ever be accountable), #530, #395