After keys are generated, the PCT are run, verifying the keys behave as expected and are usable. If they fail, the RSA calls fips_signal_error(), which moves the FIPS module to STATE_ERROR (recoverable), while for ECDSA keys, the log_fatal() is executed, which moves the FIPS module state to STATE_FATALERROR (non-recoverable).

This is not a big deal in regards to the FIPS according to the lab, but it would be great to sync on one behavior. Using the RSA approach sounds cleaner so this is my proposal