Hey,
we are also seeing this issue with our 5.2.2 cluster, we have to restart some misbehaving nodes every couple of days at the moment. We managed to bring that number down to a few nodes per week by limiting the query cache to three percent.
Our most recent node crash looks like this (you can clearly see the query cache ramping up until the node is stuck):
There is also a thread from people that had similar problems, in their case it is related to a bug in aggregations:
Hope this will help you until everything's fixed 