Cloud SQL PostgreSQL 17: SCRAM verifier exposure during password changes

I’m looking for clarification on how Cloud SQL for PostgreSQL 17 protects sensitive SCRAM password verifiers during native password changes using psql \password or libpq PQchangePassword.

Our configuration has:

  • log_statement = none

  • log_min_error_statement = error

  • log_parameter_max_length_on_error = 0

  • track_activities = on

  • pg_stat_statements.track_utility = on

  • cloudsql.enable_pgaudit = off

Native password handling avoids plaintext passwords in SQL statements, but the generated SCRAM verifier is still sensitive.

I have three questions:

  1. Does Cloud SQL redact SCRAM verifier literals in password-changing statements recorded during ERROR, FATAL, or PANIC conditions?

  2. Can pg_stat_statements, activity monitoring, Query Insights, or other Google Cloud diagnostics retain those verifier literals independently of ordinary PostgreSQL logging?

  3. What supported session-scoped protections are available to Cloud SQL PostgreSQL administrators when the relevant superuser-context logging parameters cannot be changed through ordinary session permissions?

I’m looking for documented guarantees, limitations, or a recommended supported procedure that protects verifier confidentiality without disabling instance-wide auditing.

Thank you.