Files
orca/tests
Neil ce26cd3857 test(skew): hand the two paired suites to the PRs whose claims they prove
This PR's own disposition note asked for exactly this: the suite commits
"should ride along with whichever stack PR carries the retirement-proof and
worktree-identity work". Both now do, so carrying copies here is duplication
rather than coverage.

  cross-version-session-tabs-retirement-proof -> #19860, which carries
  `publish a terminal retirement proof on the exit's own evidence`. Measured
  there: 8/8 on the branch, 5/8 red on main.

  cross-version-worktree-identity-downgrade -> #20057, which carries the
  worktree-identity re-key widening. Measured there: 6/6 on the branch,
  2/6 red on main.

Both blobs were byte-identical across the three branches -- ad44f1203c and
6bd21f1ee8 -- with no shared ancestry, so whichever merged second would have
hit an add/add conflict on a file identical to the one already present. That is
the worst kind to land on a reviewer: trivially resolvable, and indistinguishable
at a glance from a real divergence.

Removed by a new commit rather than by dropping 6a03522de1, because that SHA is
cited outside this branch and rewriting it would break those references.

What remains here is cross-version-on-disk-downgrade, which exists nowhere else:
it drives both builds' real Store over a real orca-data.json, closing the limit
the migration-pair suite states about itself -- that it measures the re-key
function, not the write path.
2026-09-11 00:39:36 -07:00
..