Having noise on the regulated power rail of your board might come from the power-supply and/or the power mains. Make sure that your motor does not feed mack currents / voltages back to your main supply.
What are the levels of the noise you see, frequency, magnitude, etc.. ?
Onboard regulators have a great deal of noise rejection themselves, so
you might suspect single turn magnetic coupling from large motor currents.
Any pins at high Z e field copupling another possibility.
Layout, are your photo couplers laid out in such a manner that AC noise isolation
is compromised, like traces coupling to each other.
Large dI/dT affects in ground bounce affecting board ? Is board galvanic from
motor ground or do they share a common DC/AC ground ?
When you designed your circuit so a stopped PSoC lead to a short circuit in your inverter, you did something wrong. You should design it so this never happens.
At least you should incorporate a watchdog to restart the PSoC automatically.
The period used in a watchdog is generally much longer than safe area
performance of motor drivers, so real solution is thermal and current
limited driver parts that will respond in uS to fault conditions.
Hli is also correct, design UP side for fault situations, although as stated
prior some conditions need to be taken care of by device level protection.
One thing you can do is to take the outputs of the PWM (I presume you have two because otherwise you won't run into the problems described) and lock them (inside the PSoC) so only one can be active at a given time:
(assuming both outputs are active high)