Instead of debug logging, have the scheduler record all messages to a file. Also take state snapshots of active fsms in between ClientRequest messages. This can be used as a human readable trace.
Record the shrunk failing ops and implement a method to replay them.
Make replay accessible to gdb
Ensure quickcheck shrinking works as expected
Enhance quickcheck tests to:
Interleave Client Requests
Shuffle (in a causally consistent manner), scheduler buffered messages.