it seems the script aborts after the first ssh error if the retry_limits are not defined

Logic LGTM, but as written --reboot-retry-limit doesn't seem to have to do anything with the actual reboot (into the installed system, at the very end), but just retries connecting after kexec. While kexec could be considered a form of reboot, I think this naming could confuse users and would suggest something like --kexec-connection-retries or so