As a follow-on to discussion here

We lose test coverage because in a failover scenario, we do not restart index builds when the new node steps up (rather, we abort it with InterruptedDueToReplStateChange). This is ok for the user, as drivers retry upon getting this error, but it's different behavior than tests are expecting.

If we emulate this behavior either via a suite override or some other test adjustment, we may be able to re-enable some or all of these tests.