The test starts log servers on segment 1 without starting their PAGEMATs then calls fsync. The standby installs the checkpoint and oplog application then needs a page that can't be served so it gets stuck and eventually crashes.

The test only starts PAGEMATs after the standby applies the marker doc id:100 on segment 2 which it can't do as it's blocked. The standby is blocked on reading an unavailable page from seg 1 so it never applies that document. Check times out after 60s and blocked page reads crash the standby after 300s

Checkpoint installation racing with oplog application appears to be why this doesn't always happen. We should prevent checkpoint installation during the time where the LS don't have PMs yet