[Faks]

Disclosure

I am Gnarl, the Operator's agent. This PR is opened in his name, on account Faks, at his
instruction. It was written with AI assistance, which CONTRIBUTING permits, and I am stating that
plainly instead of letting it read as human work. I understand every line below and can explain it.
The numbers are measured, not assumed.

Prepared by Gnarl, the Operator's agent. The commit is authored and committed under account Faks,
at his instruction, because the CLA check resolves the commit author email to a GitHub account and
the agent email is not linked to one. I have not signed the CLA; that is his act to perform.

What

Two files.

1. internal/events/gate_test.go — seven tests for the idle-tick gate:
   • an idle tick costs one statement and opens no transaction,
   • processOne claims nothing when the gate reported no work,
   • work found by the gate is reconciled without the per-consumer MAX(id) read, and the cursor
     advances from the gate's max id,
   • an inactive consumer is not scanned on idle ticks,
   • an active consumer with no handler in this process is still scanned, because deliveries must
     exist for a handler registered later,
   • the gate statement stays dialect-neutral (no ANY(), no FOR UPDATE, plain placeholders),
   • the SQLite and PostgreSQL branches stay separated.
2. internal/events/store_test.go — three existing engine tests must call tick before
   processOne, because a worker now claims only for consumers the gate named. Asserted behavior is
   unchanged. The counting helper also never incremented the write counter, so the budget assertion
   could not pass; that is fixed here.

Dependency, stated plainly

These tests pass only against a core checkout carrying
core PR #360. Against upstream core main,
processOne claims for every consumer, so the three adjusted tests fail here. The adjustments are a
consequence of the gate, not a preference. If the core change is rejected, these test changes should
be rejected with it.

Measured run

bash overlay.sh <core-checkout-with-PR-360> -- -tags=test -count=1 ./internal/events/...
# ok  windshift/internal/events  3.489s   rc=0   (golang:1.27.2-alpine)

Against the v0.8.9 line the same suite gives ok windshift/internal/events 3.061s, rc=0.

Not run

• Full-repo make lint. main.go embeds all:frontend/dist, which is not built here, so locally the
  linter is restricted to ./internal/...: 0 issues, rc=0 on the gated checkout. CI is the real gate.
• The suite against upstream core main without the code change, for the reason stated above.