This one is a little hazy, sorry about that.
A few weeks ago, when CAC was unhealthy, I observed a scenario where a particular page-diff request was running for several (15+) minutes. Unfortunately, because of the bad health of CAC at the time, I was unable to get a heap-dump or determine exactly what URL was requested that lead to this scenario.
What I do know is that the thread was executing for 15+ minutes, and for the entirety of this duration was stuck in org.eclipse.compare.internal.core.LCS. firstname.lastname@example.org has indicated that this should never happen, as page diffs have timeout logic implemented. Is it possible there's some edge-cases, perhaps a page diff that is extremely computationally expensive, that cause the request to exceed its timeout and not be properly interrupted?