After updating our application code and moving to pymongo 3.0.3 in production we are seeing a jump in load, context switches, local timer interrupts and number of threads. I've attached a couple of screenshots from Munin that shows a drastic change in resource use on one of our application servers right after the deploy. We are still investigating this, but as nothing else was patched at the same time things points to new MongoClient behaviour.

The application is a collection of web services run through gunicorn.