We were performing a series of restarts of config servers and shards. While some of the servers were rebooting, mongos that used them occasionally returned the following error:

).
Our auto-retry framework relies on a set of error codes and/or messages and can't cope with such errors. Emitting a more descriptive error and assigning this kind of error an error code (even generic one) would be very helpful both for diagnosing reasons of the error and attempting to recover from it.