The request basically asks to run a regex against 76M titles and fails with timeouts. While the error messages could be improved, this is such a niche thing that I don't think it's particularly important. This happens to fail the timeout in a different way than other things, but the only improvement i would probably be able to offer is to replace the whole search results page in this use case with a message that effectively says "dont do that"

This does not timeout but the number of results is also wrong:
https://www.wikidata.org/w/index.php?sort=relevance&search=haswbstatement%3AP31%3DQ6256+intitle%3A%2F%5Ba-z%5D*%2F&title=Special%3ASearch&profile=advanced&fulltext=1&advancedSearch-current=%7B%7D&ns0=1&ns120=1&ns146=1

Actually there's 222 items with P31=Q6256, but Qid of none of which contains only letters.

While it doesn't say timeout, the request spins for some time and eventually the backend gives up and reports a failure. As i said the error messages shown to users could be improved, but the request would still fail.

Hmm, the request with only 222 items is interesting, will look at that

Looks like it's hitting this: https://gerrit.wikimedia.org/r/c/search/highlighter/+/435282/5/experimental-highlighter-lucene/src/main/java/org/wikimedia/highlighter/experimental/lucene/hit/AutomatonHitEnum.java#189