This could lead to wrong execution scenario like the following:
- Collmod starts, stop migrations and enter the kUpdateConfig phase
- Hit a retriable error and unblocks migrations
- Attempt to re-execute the kUpdateConfig but this time with the migrations unblocked
Keep in mind that we can't simply resume migrations on non-retriable error, in fact even after hitting a non-retriable error we can't guarantee that the coordinator won't be recovered and re-executed from a new primary node in case of stepdown.