log_lock_waits

With log_lock_waits on, any session that waits for a lock longer than deadlock_timeout gets a log entry naming the lock and the blocker. It's the go-to setting for diagnosing contention and 'why is this query hanging'.

Turn it on in production — lock waits are a common, hard-to-reproduce source of latency. It piggybacks on deadlock_timeout (default 1 s), so it only logs genuinely stuck waits, not every brief one.

• deadlock_timeout
• lock_timeout