It is not banned directly, but the hardware is not [let's call it] "thread save".
Two examples where you get problems with two cores:
- if you want to prevent the device from getting stuck in the code -> perform reset if not responding. If core0 gets stuck, core1 will still reset the WDT and no reset will be generated.
- If you want both cores to trigger on periodic interrupts generated by the same WDT and one core is busy in a higher prioritized ISR and the other core clears the Interrupt flag, WDT ISR will be missed on the busy core.
The second one might be solvable by a semaphore like construct, where you wait for the second core to enter the ISR as well.