Today, execution control tunes the number of available tickets, but it doesn't do anything to limit the number of operations queued waiting for a ticket. We rely on the operations timing out or being interrupted externally to remove themselves from the queue once it becomes prohibitively long/operations can't acquire a ticket in a reasonable amount of time.

It would be better if the ticketing/queueing system instead had a means to directly limit how many operations could be waiting for a ticket. We should add a tunable/runetime-settable 'max queue depth' parameter for the different ticketholders, and return a new "MongodbOverloaded" type of error when an operation attempts to acquire/wait for a ticket but we're already at the max queue depth.