Id personally go DB and store previous run time maybe even start/finish in a table. Bonus is you get to see how often people are kicking off the process. Tho I am in the boat of never manually kicking off scheduled processes. Expanding on this, we normally go off the start time and make sure the job hasn't run for at least x minutes which is normally double normal run time. In case process exploded and never recorded an end time. We also track wether it was a system called job or if a user kicked it off manually based off token security so we can ask why someone thought they needed to manually.

Yeah, that was our first though (DB). It seemed like less work to not involve the DB, but maybe it is worth it.

We use DB because we have multiple servers so app scope just wasn't a good use and honestly it isn't persistent if we reset the server

Yeah, that makes a lot of sense.