Flaky: runtime_watch_snapshot_is_an_error_when_the_store_is_unreachable races on a freed port#4146
ployzd corrosion::store_tests::runtime_watch_snapshot_is_an_error_when_the_store_is_unreachable fails intermittently in CI (rust-tests daemon-2, PR #4143 at 588698bb74):
| 1 | panicked at crates/ployzd/src/corrosion/store_tests.rs:180:5: |
| 2 | assertion failed: store.certificate_rows().await.is_err() |
The test binds 127.0.0.1:0, reads the port, drops the listener, then expects nothing to answer on it. Under nextest another test can bind that freed port in between, so the "unreachable" store answers. It passed 5 of 5 locally and on the previous CI run of the same PR, and the PR touches nothing under corrosion/.
Fix idea: keep the address unreachable by construction, e.g. a port held by a listener that never accepts, or an address that refuses (such as a closed socket the test keeps open in a state that resets connections), rather than a freed ephemeral port.
F 3 (one CI leg, now and then) × S 2 (a re-run fixes it) = 6, so filed rather than fixed in #4143.