Okay yes, I see from the logs that there have been a bunch of them. The amount of resources for the services might be in need of adjustments still, but let's see if that was the fault of the slow background calculations.
edit: Yes, most probable cause was that the request timeout was set to unnecessarily long time per each user request, and since GW#2 database was glogged the user page load requests filled the quota, giving occasional timeouts. For now I have not seen those any more after some adjustments were made. But please let me know the page and exact time if it happens again.