Hello, @yashasdu
Sorry for the late response.
Since I cannot reproduce the issue from my side, after recent days investigation, a few suggestions provided for your further debugging on the custom board.
1. Comparing with legacy reboot and panic reboot, there are not much code differences from kernel perspective, the main difference is that no do_kernel_restart_prepare() is called via the panic reboot flow.
2. While the execution flow left the kernel and switched to the TFA, the actual operation for both legacy boot and panic boot is the "destructive reset", there are no differences of the two operation from the low level.
3. Not sure if there are completely no issues of your legacy reboot tests, further checking your reset schematic may also be a option for further debugging on the issue.
I feel sorry that I am not able to provide directly debugging support on the issue since it cannot be reproduced with RDB3, apologize for your inconvenience.
BR
Chenyin