The metamorphic test uses MemFS.CrashClone in the dbRestartOp, but that isn't concurrent with an Open. Similarly, TestCrashOpenCrashAfterWALCreation crashes after the Open returns. We don't seem to have code coverage for a crash during Open.

We could add another operation to the metamorphic test that does the unclean restart like dbRestartOp but also starts a short-lived concurrent goroutine that calls CrashClone during the Open and uses that clone do a second Open. We would inject slowness into the FS for the first Open, so that it is slow enough that we capture some arbitrary intermediate state.

Jira issue: PEBBLE-342

Epic PEBBLE-1131