After much experimentation and such, I think I've solved my comprehension problem. Essentially, I've been bringing a PostgreSQL (PG) mentality to this situation and that has prevented me from seeing the bigger picture here.

Here's what I've learned:

• The kill* operators in MongoDB do indeed kill sessions, operations, etc. correctly. I confirmed this mostly with my killOp() experiments.

• In PG, you can kill "client sessions" off (i.e. drop their network connections to the DB). It is a necessary step when dropping SQL DB's for example. MongoDB client connections (e.g. mongosh, a pymongo session) are not tied exclusively to a particular database and so MongoDB has no reason to require killing client connections for dropping a DB.

• MongoDB clients like mongosh and pymongo have logic inside them to keep things alive and so killing sessions and operations might not make a visual difference on the client side.

Thanks to everyone who helped me.