Regarding the migration procedure the end state should be aimed at being fully quantum resistant, carrying through quantum vulnerable primitives like ECDSA does introduce future liabilities and shovels the problem forward.
Notably, legacy software will still be vulnerable if an attacker controls validators, in which case a fork would be needed. If a fork is needed anyway it would be more advisable to introduce quantum resistant signatures directly in the codebase in a way that offers unconditional security and maintains the integrity of passphrase standards.
Half measures generate a false sense of security and does not incentives a complete upgrade.