Answer
IFSSYS specifically requires recreating each clustered node after its password is changed — simply running the Installer's password-change option on one node isn't enough for IFSSYS in a clustered configuration, and skipping the node-recreation step for the other hosts results in the IFSSYS account getting locked out. Other internal accounts can behave inconsistently in a cluster too (for example, an installer-driven database password update for one account silently failing to actually take effect in the database despite the option being selected, requiring a manual database-side password change plus a standalone installer rerun scoped just to that account) — so after any internal-account password change in a clustered environment, verify each affected service actually reconnects successfully rather than assuming the change fully propagated.