This stems from the fact that "can't find self in new replset config" is triggered if a member doesn't find exactly one "self" in the config.

As such, if there are multiple entries it can cause problems, and each mongod process that is a member of a replica set caches all known self-hostnames they've learned about since the process started.

Therefore a solution is needed to determine what the contents of this cache is (or to clear it) so that issues that trigger "can't find self" can be predicted and circumvented