We have a large site of 30,000+ users using multiple web servers one of which is entirely dedicated to cron.
We're finding that ad-hoc tasks are really piling up. At the time of writing the queue is 90,000 items long and about 15 hours old (for the oldest entry). This isn't very good.
I'm inclined to increase the task processing settings but - even though I think I understand what they mean - have no real feel for these figures at all.
The machine running the cron is showing very low load but, of course, there are consequences for the DB as well.
Has anybody got any useful input?