This is to address the fresh flakiness in the 4.0 branch of the change_stream_shard_failover.js test, which fails the chunk migration during failover. The failure logged at client side:
"moveChunk command failed on source shard. :: caused by :: NotMaster: not master"
I think logging this at replica side to detect where and when the migration command failed would help.

The flakiness is very recent (since May 28) and we should understand if recent changes triggered it. I cannot reproduce it locally, never fails.