I say it needs more than a comparator's hysteresis - any spike or noise can still false trigger that. Any filtering introduces a lag.
As well the problem is 3-phase transmission and distribution systems have failures such as loss of a phase, phase-phase short etc. as lines get tossed in the wind, that squirrel gets roasted, one of three fuses blows etc.
Breaker reclosing and tap changing voltage regulators also cause disturbances up to a few cycles.
If a phase drops out, gets lost, loads will backfeed and put voltage on the lost phase. I had one site missing a phase (broken wire at pole transformer) and the induction motors were not happy, growling a bit and through transformer-action they were putting voltage on the missing phase. You don't measure 0VAC. It's just a low voltage and the angle was lagging, way off. Power meter read -ve power for that phase on a motor, +ve high power on the other two. Had me totally confused.
Search for phase sequence detector or phase-loss detector these are used to protect motors and there are tons of weird circuits out there.
Yes you apparently have a resistive heater but what will it do with a missing or shorted phase?
If you have 3 zero-cross detectors, one for each phase, you still need to detect zero-sequence.
So all this ends up best implementing a PLL or oscillator in firmware.
The MCU has a sync'd oscillator and knows mains angle internal and compared to what is seen. You can wait for several seconds before syncing. Then a small correction for drift and small frequency changes.
Because mains absolute phase does not change unless there is a blackout or grid collapse, so the MCU will have an accurate mirror of it.