Single-VPS vs multi-region: when to scale

SwyDex runs production on a single VPS. We hear “but what about HA?” about once a week. Here's how we think about it.

Cost. A multi-region setup with proper failover is 4-5x the bill of a single instance.
Operational simplicity. One Postgres, one Redis, one app process. We can debug everything in one terminal session. New engineers ramp up in days, not weeks.
Latency for the dashboard. The dashboard is a single-region app for our customers. They don't need DNS-based geo-routing.
Data consistency. Multi-region means multi-master or careful read-replica routing. Either is a footgun.

When customer concentration in a different region passes ~30% — latency starts to hurt.
When our single-VPS uptime drops below 99.9% — currently at 99.96% on the rolling 90-day window.
When a customer's contract requires it. Enterprise customers sometimes do.

Off-site nightly backups (encrypted) to a separate provider.
Documented restore procedure that brings up a fresh VPS from backup in < 2 hours.
DNS TTL of 60 seconds so we can swap origins quickly if the VPS goes down.

Single-VPS is a deliberate choice, not a fallback. It'll change. Just not before it has to.