Sure, you can overlay fail‑safe layers on a deterministic core scheduler, but you have to keep the critical path isolated. Use two copies of the state machine, sync them on a separate bus that doesn’t carry the real‑time data, and drop any backup processing that could introduce jitter. Treat the backup as a “shadow” that only kicks in on a fault, not a regular path. That way the timing stays clean, and you still have redundancy if something goes wrong.