We have seen too many instances of slow machines or threads that take > 30s to perform work while holding the RSTL, crashing the server can be expensive since restarting is not a quick operations, we think it will be better to give more time for threads to finish their work to avoid crashing as often.

Ideally we would be better at interrupting work, but that is a much larger issue. We think the negative change of an actual deadlock taking 60s to realize vs 30s is much less than the negative value that crashing the server with no deadlock brings