test/format currently does crash recovery testing or not crash restart testing, but it only ever does a single restart. A recent bug (WT-10551) required multiple restarts to reproduce in local testing. We should explore having test/format restart multiple times against the same database, to make it more likely to uncover similar bugs.

The original bug involved a backup not capturing all data, so some form of correctness checking is necessary. i.e: playing forward across restarts and in parallel playing forward without restarts. Then validating that content is the same in both databases after.