Replacement of the unhealthy etcd member doesn't work with node being automatically rebooted

Issue

Trying to follow the official documented procedure to replace a failed ETCD member on bare metal IPI clusters makes the cluster almost completely unavailable and impossible to proceed with the replacement once the master is shutdown or removed on the first steps.
Seeing nodes being rebooted automatically which disrupts the entire process.
Deleting selfnoderemediations CRs doesn't work as they keep getting recreated.