in BF-20634 we have an unclear cause of failure. It could be one of:
- dropRole could have failed, and since we are not checking result we can't tell it didn't
- call could have returned before replication completed, and since we are not checking primary first, we can't tell it's not the case
I am making an attempt to reproduce an issue on windows/debug, so far without success.
After discussion with xuerui.fa we decided following needs to be done:
- check that dropRole succeeds
- check that after dropRole succeeds, the role has indeed disappeared from primary before checking all other nodes
This will not address the root cause, but will help eliminate possibilities