Catch error spikes and stuck loops

This scout reports error spikes and ranks them by how many guests they affect. A bug that breaks booking for hundreds of guests ranks above a few browsers stuck in a retry loop.

What the scout looks at

The scout reports an issue when its exceptions jump well above the issue's own baseline. It sets the priority by comparing exceptions with distinct users:

  • Both climb. Many guests hit a regression. The report is P1, or P2 when the reach is moderate.
  • Exceptions climb, users stay flat. A few browsers are stuck in a loop Fig. 1. The report is P3, or P2 when it's fresh and in the thousands.
  • A resolved issue fires again. The report is P1.
02:10

Distinct users

flat at 4

Exceptions

~20/hour → ~240/hour

Fig. 1 – What the scout saw: At 02:10, exceptions on Twig's Trip details issue jumped from about 20 an hour to about 240 an hour. Distinct users held at four the whole time.

The scout files new issues with overlapping stack traces as one report, and skips one-off exceptions and known upstream errors.

The report in your inbox

Each spike becomes a report in the team's inbox, with exceptions, distinct users, and an hourly chart against the issue's baseline Fig. 2. From the report, the engineers can ask AI who else is affected or what to do next.

P3Trip details retries loop on the same TypeError, ~240 exceptions an hour from 4 guests Exceptions jumped to ~240 an hour at 02:10 while distinct users held at 4, because four guests keep clicking Try again on a Trip details page that throws every timeScout · error trackingReadyActionable3h ago
ArchiveReview
Fig. 2 – The report card in your inbox shows the scout's claim and triage. The full evidence is inside the report.Hover over the figure to learn about each element.

The scout itself

PostHog maintains this scout, listed as Error tracking in the project's scouts. Its skill file is on GitHub. Each run queries Twig's issues and $exception events through the PostHog MCP, and checks its notes and the team's inbox so it doesn't report the same spike twice. It runs every 24 hours by default and stays quiet unless something needs the team's attention.

Add this to your scout troop

The engineers turn it on for each project. It needs exceptions reaching PostHog.

Open your scoutsFind Error tracking in the list and switch it on.