Immediate alert or silent retries first?

Silent failures are the absolute worst especially when users rely on sync data for daily workflows. But spamming someone over a temporary 5-minute API downtime is just as bad.

I usually prefer a quick retry loop behind the scenes first. If it’s still broken after an hour, flag it in the app with a timestamp.

Though if it's a dead auth token that probably needs an instant heads-up since retries won't fix it anyway. How do you guys decide that threshold?