Yes, we spotted that contradictoin betweeb 'crash every 15 minutes' and 'suspect code only runs 20% of the time' later. We still don't entirely understand it.
Since we commented out those three lines of code, the servers have been running fine. Derek (Woolhead) is trying to work on a more long-term fix. If we get one, we plan to submit it for integration.
I am more interested by the 0.35s -> 0.5s increase in the average load times. I have no idea what is causing that.