Darling's deadlock maintenance logs nothing when a stage recovers after a failed pass #5634

Claude posting for Erik Darling.

Darling's hourly PostgreSQL deadlock maintenance logs a warning when one of its stages fails a pass. For example: "failed 2 of 4 ... the next pass is skipped". When a later pass of that stage succeeds and has nothing left to do, it logs nothing. The failure count is reset without a word.

So an operator who reads the warning has no log line that says the stage recovered. Seen on a large store on 2026-10-09: the last line for the stage was the "2 of 4" warning. The retry about two hours later succeeded without a log line. The only way to know was to read the code.

Fix: when a pass succeeds after one or more failed passes, log one information line naming the stage and how many passes had failed. Code: PgDeadlockRemask.cs, where a clean pass resets ConsecutiveFailures.