The Execution Loop

Your scheduler's main loop:

while running:
    next_run, job = heap.peek()          # earliest-firing job
    delta = next_run - now()
    if delta > 0:
        sleep(min(delta, 60))            # cap at 60s for responsiveness
        continue
    heap.pop()
    spawn(run_job, job)
    job.last_run = now()
    job.next_run = next_run(job.cron, job.last_run)
    heap.push((job.next_run, job.id))

The 60-second cap matters: if a job is added with a sooner firing time while we're sleeping for an hour, we'd miss it. Some schedulers wake on signal (condition variable) instead.

spawn(run_job, job) should NOT block the loop. Use threads, asyncio, or spawn a subprocess and wait() separately.