Possible regression of #8815 ("Nordic: Directly accessing GPIOTE might create unstable firmware"...)

[JoeHut]

With Zephyr 1.13 I ran into a similar issue as described in #8252, which was actually fixed in #8815. After a certain point I was unable to receive interrupts on a pin but could confirm with an logic analyzer that the state had changed. Similar to last time the issue was only seen when PWM was used as well. As a workaround, I switched to the hardware based nrfx pwm driver which solved the problem (or at least it is hidden now). I suspect that the rewriting of the nrfx-based gpio shim in combination with the old pwm_sw driver reintroduced overwriting of GPIOTE channels.

cc: @cvinayak

[JoeHut]

Fixed in #11807