Should the on-call person get every alert or only critical ones?

One in five is a strong signal that the routing, not the monitoring, needs work. A structure that works well: 1. Split your monitors into two groups by consequence, not by system: things that need a human tonight, and things that need a human this week. 2. For the first group, subscribe the loud channels. SMS and voice calls, routed to the on-call person, with a full escalation chain behind them. 3. For the second group, subscribe only quiet channels. Email or a Slack channel is enough. Do not attach them to the rotation at all. 4. Take staging and other non-production environments out of the on-call path entirely. If a staging outage cannot lose money or customers overnight, it should not be able to ring a phone. 5. Review the split every month. Any alert that was paged and not acted on within an hour is a candidate to move to the quiet group. One technical detail that helps you here: notification channels are subscribed per source. That means you can attach completely different channel sets to two different monitors without any global rules. That per-source subscription is exactly the lever for this problem.