A standby can replay WAL from a primary, but replication doesn't replace independent backups. A destructive statement can replicate too. Keep retention and recovery options that survive an application mistake or accidental deletion.
Asynchronous replication can lag. A user who writes on the primary and immediately reads a standby may not see that write yet. Route consistency-sensitive reads to the primary or use a measured replay-position strategy. A time-since-last-replay metric alone can mislead on an idle primary because there may be no new transaction to replay.
Synchronous replication changes commit durability and availability tradeoffs according to the configured acknowledgment level and standby set. Failover also needs fencing so the old primary can't continue accepting writes. The browser has no standby process; these tasks assess supplied operational evidence and the boundaries of what it proves.
Asynchronous replay can make a successful write temporarily absent on the standby.