When cron itself is running fine, but cron is congested it should warn in the check. This happens when you have for example 3 cron runners, and all are tied up on tasks which take a few minutes. So each minute nothing new can start.

So proposing to detect this state by looking at the last time a new task started, rather than the time the cron cli process ran. It may need a congestion logic for both scheduled tasks and adhoc tasks individually.

Bonus points: a completely different way of viewing this is to compare what need to run, vs how much we know we are processing. Something like 'we know we have 6 cpu hours of tasks to run in the next wall clock hour, but we are only running 2 hours worth, so we are not keeping pace.'. This would be a 'smoother' better metric if it is viable