Currently, the edit history page for large collections (on the order of a few thousand releases) will time out and generate an ISE until the cache is warmed up for that query. As discussed on IRC, I did some profiling on this around the time it was implemented and the issue appeared at that time to be in the join to the edit table. As Ulrich pointed out, the query has since been modified to improve the query plan for this join, but the timeout issue persists.