Category

Engineering

Notes on how StatusLoom is built.

Engineering
Why We Wait for Two Failures Before Calling a Site Down
A single failed check is usually a network hiccup, not an outage. Here is the reasoning behind the 2-strike rule, and what it costs.
Aug 23, 2026 · 2 min read
Engineering
Five Ways a Check Can Fail, and Why We Tell Them Apart
Knowing a site is down is half the information. Knowing whether it is a DNS failure or a timeout is the half that tells you where to look.
Aug 8, 2026 · 2 min read
Engineering
Why Every Timestamp Is Shown in Your Own Timezone
An alert that says a site went down at 2:20 AM UTC makes you do arithmetic before you can even understand what it is telling you. So we do the arithmetic instead.
Aug 3, 2026 · 2 min read
Engineering
Recovery Needs Only One Success, Not Two
Going down takes two consecutive failures. Coming back up takes just one success. The asymmetry is deliberate, not an oversight.
Jun 24, 2026 · 2 min read
Engineering
Partitioning the Checks Table: Keeping 90 Days of History Fast
Every check, for every monitor, on every interval, adds up fast. Here is how the checks table stays fast to query without keeping data forever.
Jun 19, 2026 · 2 min read
Engineering
The user_id Column Is the Tenancy Wall: Single-Tenant by Design
There is no team model, no invite flow, no shared monitors. One account owns its monitors outright, and that simplicity is deliberate.
Jun 14, 2026 · 2 min read
Engineering
Inside /ping: The One Route That Doesn't Need You Logged In
Almost every route requires an authenticated account. One doesn't, and it exists for a very specific reason.
Jun 9, 2026 · 2 min read