The story read counter is off again, this time for good or until the feature is re-implemented. Updating a big and very busy table every few seconds just doesn't work well with MySQL table locking stragegies. Every now and then, an story read counter update statement took too long, blocking all select statements on the table, driving up CPU usage and response times.

I guess the solution would be to use a separate supersimple table for story read counters.

(nevermind)