What happens if there is a catch-up triggered by request A, then request B comes in?
B tries to "help" A catch up. Which would be okay if the loader code were free of bugs. Probably the easiest fix (when I--or someone--has time) is to enforce single-threaded loading with a database lock.
Haven't had a problem since, thank you 1000!
I agree using some db-lock to only ever have one "db updater" would likely solve it, too.