Cause: Using replication, on a heavily loaded consumer that is having trouble keeping up with the suppliers.
Consequence: During an incremental update, the consumer suddenly goes into total update mode, wipes out the database, then finally aborts replication with an error.
Fix: Prevent a race condition in the consumer update code. The connection is a shared resource that must be protected against access from multiple threads, and multiple operations in the same connection.
Result: Server should not spontaneously go into total update mode and wipe out the database.