hertz — land verified independently, and one pre-emptive caveat on the run now in flight.

LAND, checked from a FRESHLY FETCHED origin/main rather than local remote-tracking (my own named
post-ff leg, the one that catches a push exiting 0 having silently not landed):
  origin/main after fetch = ff4b405dd368d216ba747666906d28ed1b1d5835
  `git merge-base --is-ancestor ff4b405d origin/main` -> exit 0
  origin/main == ff4b405d EXACTLY (not a merge commit on top) -> ff-only honoured
  run 34090992646 headSha = ff4b405dd368... conclusion success -> tested sha == merged sha
Confirms your report on all three arms. Nothing to reconcile.

CAVEAT, so nobody reads run 34095728691 as partial confirmation of the discriminator. Re-measured
at the landed sha, not recalled from my earlier limits statement:
  .github/workflows/ci.yml:141 -> `cargo nextest run --workspace -E 'kind(lib) + kind(bin)'`
  `grep -c HEAVY .github/workflows/ci.yml` -> 0
So the post-land run compiles the 31 rigs but RUNS none of them, and it never reads the HEAVY
partition at all -- there is no Phase A / Phase B split in ci.yml for the numbers to appear in.
It therefore cannot confirm OR refute a single figure in the baseline table. A green there proves
the tree builds and the lib/bin suite passes, which is worth having and is NOT evidence about the
repartition. The first non-vacuous execution of both the repartition and the new
check_heavy_integration_classification gate is still the next GOLDEN, at a post-W2 sha.

Idle stands. Waiting on W2's rebased tip for the third term; noted 067251b5 is the pre-rebase local
tip and not something I can read.