A fixed table (MCS_SCHED_MAX_JOBS = 8) of jobs. A job is a script file (from the VFS) or a C# delegate. The scheduler owns no thread: the host calls mcs_sched_poll(&s) from its loop/task; due jobs run there one after another as normal top-level runs, so the execution limits, hook and error reporting apply to each job. mcs_sched_poll returns the milliseconds until the next job is due (sleep that long) or -1 when nothing is left.

mcs_sched_t s; mcs_sched_init(&s, vm, &vfs, millis, NULL);
mcs_sched_open_lib(vm, &s);                       /* C# Scheduler class */
mcs_sched_load_config(&s, jobs_cfg_text);          /* optional jobs table */
for (;;) { int32_t ms = mcs_sched_poll(&s); rtos_sleep(ms < 0 ? 100 : ms); }

# kind    time   path            policy
startup          /init.cs
after     5s     /selftest.cs
every     500ms  /blink.cs       restart=always
every     1m     /report.cs      restart=3

startup /main.cs is skipped at boot: the shell runs /main.cs anyway (it would run twice).

<time> = number with optional ms/s/m/h suffix. restart=never (default: stop after the first failure), always (never stop), or N consecutive failures. Periodic jobs that fall behind run once late, then continue one period later (no burst catch-up).

int id = Scheduler.Every(1000, () => GPIO.Toggle(13));   // optional 3rd arg: maxFailures
Scheduler.After(250, () => Console.WriteLine("once"));
Scheduler.Cancel(id); Scheduler.CancelAll(); int n = Scheduler.Count;

Delegates are kept alive by one pinned list per scheduler.

The VM is single-threaded (with an OS chosen, file jobs run on their own OS threads unless marked main — THREADS.md), but due jobs also run while a script waits in Thread.Sleep (MCS_SCHED_DURING_SLEEP, default 1), so a main loop and jobs work together:

Scheduler.Every(1000, () => Console.WriteLine("tick"));   // or `every 1s /report.cs` in jobs.cfg
while (true) { /* poll sensors... */ Thread.Sleep(50); }  // jobs run during the sleeps

A loop that never sleeps starves them. A job that sleeps does not start other jobs. A script job (jobs.cfg) that runs inside a sleeping script gets its own top-level variables: the outer script's values are restored afterwards (mcs_globals_save / mcs_globals_restore).

Cron expressions, wall-clock/RTC schedules, priorities, preemption, job persistence across reboots (other than re-reading jobs.cfg at boot).