The ONLY times I have seen this happen is shortly after a bounce (whether deploying something or not). Usually it’s only 2-3 instances within a few hours of traffic after the bounce. But after bouncing the impacted instances again, then it seems to be resolved. Not sure if there is some condition while recompiling after the bounce when users hit it that calls the hangup or what the deal is. In this case, it looks like the instances were bounced 11/27 at around 3:30am and the issue happened at 10:30am the same day (2.5 hours after start of business).
Joined: Dec 03, 2011
If possible, post the stack trace at the time of issue occured i.e. 11/27 around 10:30am.