@grapevine , again, it can be very helpful when you provide additional base information, such as the actual physical devices being used and the IOS version being used. (Sometimes it helps others find a documented bug that covers the issue you're encountering.). Further, answering any specific requests for specific information.
So far, you've confirmed the ping loss, and drops of Etherchannel links, on FW, when stack master is powered off (?).
From what you describe, the impact of loss of stack master seems more impactful than I would have expected. Assuming a bug is not involved, I would suspect, possibly, the configuration isn't optimal. I have in mind SSO (I believe this is mostly/fully "automatic" for a stack), NSF (often needed for L3 protocols, like routing, to keep neighbor relationships active during fail over recovery - also may require neighbors to be NSF aware [FW?]) and possibly LACP parameters too.
From your description of ping drops, you didn't describe whether ping loss was for stack as source or destination, or all ping traffic was transit, or different results for transit traffic vs. local src/dst.
The big issue appears to be Etherchannel links seen going down on FW. I've no experience with a Palo Alto FW, and possibly, it and a Cisco physical stack might not ever play nicely together if certain physical stack member failures. Possibly, the FW needs configuration changes.
Assuming the PA FW and Cisco physical stack cannot cooperate, one or both, unfortunately, might need to be replaced. On the Cisco side, possibly a Stackwise Virtual might work with the FW.
Have you contacted either/both Palo Alto and/or Cisco TAC? What you're trying to accomplish isn't oddball and both vendors should have more experience with this kind of requirement.
