Could probably configure the application server to allow for a longer timeout, but also is it known why the computation is so slow? It seems like vmassol is recommending providing an error message, but what about looking at improving the performance of the diff to avoid timing out in the first place. Mediawiki does rendered diffs by default that seem both accurate and performant, so it seems like a solvable problem unless that comparison isn't apples to apples for a reason I am missing.
This message was sent by Atlassian Jira (v9.3.0#930000-sha1:287aeb6)
If image attachments aren't displayed, see this article.