No, much simpler - store a checksum alongside public keys in places like .ssh/authorized_keys, and have the software like sshd recompute the checksum of the in-memory key each time it uses it for authentication.
This doesn't sound correct. What if the attacker times the operation so that the bit corruption occurs after the checksumming but before being actually used for cryptographic operations?
The attack relies on glitching the in-memory key.