Issue #3462456 followup by o'briat, japerry: fix: probe dead servers once per request.

Drupal 11.2 runs both hook_runtime_requirements() and the legacy hook_requirements('runtime'); this module implements both to keep Drupal 10, so the status report probed every unreachable server twice. The probe result is now kept in a static for the rest of the request.

The failure log said "the site is running without this cache" for any unreachable server. That is true for a single server and wrong for a cluster, where only the keys that hash to the dead server are affected and the other servers keep serving theirs. The log entry and the status report row now say which of the two applies, and the status report row explains where the lost share of the cache is rebuilt.

Test: a second requirements run in the same request must not probe again; the helper resets the static between the simulated requests.

Closes #3462456

Merge request reports

Loading
Loading