How many mongodb connections are you monitoring? From log it seems like multiple connections are used. makes it difficult to follow or tell which OP_QUERY messages are send on same TCP connection.
Not knowing mongodb in detail yet, one of the problems producing this warning might be either a client piping requests (sending multiple requests, before seeing first response). For example in async usage? Or a message type not requiring a response not being handled correctly by correlator (but OP_QUERY requires a response).
Checking mongodb wire protocol and mongodb code correlations do not take the query its requestID into account when looking up the connection. The wire protocol seems to support out of order responses + piping is always possible.
Additional some request types (e.g. OP_UPDATE) do not have any kind of response. For example if your connection has one UPDATE followed by a QUERY you will get this message, but can be ignored.
Still would be helpful if we can isolate the connection errors happen on for verifying my assumptions. E.g. get a pcap with tcpdump. packetbeat provides a debug mode for reading from pcap (-N path/to/trace.pcap). Having a pcap and using tshark/wireshark connections can be isolated.
Please follow this issue on github