What Problems is for
A raw capture treats every failed request as an unrelated event. Problems doesn't — it classifies failures by cause and groups related ones together, so ten failed requests to the same host during a five-second outage show up as one entry, not ten identical rows to scroll past.
The categories
- HTTP errors — 4xx/5xx responses, grouped by host and endpoint.
- DNS and connection failures — the request never got a response at all.
- Upstream TLS issues — the server's own certificate failed validation, distinct from anything about your device.
- Interception — certificate pinning or handshake failures specific to how Hollowport intercepts traffic, kept separate from ordinary connectivity problems so the two are never confused.
Reading an entry
Each row leads with what happened in plain language — not just a status code — plus the scope (how many requests, how many hosts) and whether it's still ongoing or has recovered. Tap it to open the full detail: when it started, whether it's recovered, the specific endpoints involved, and, when there's an earlier successful request to the same endpoint, a direct path to Compare the two.
"First observed," not "started"
Hollowport only shows what it actually captured. If a host was already failing before you started capturing, the detail view says "first observed" rather than claiming to know when the problem actually began — a small distinction, but an intentional one.
From Problems to a fix
A Problems entry is usually the starting point for an investigation: read what's known, compare a failure against a working request, then replay the request with a change to test your hypothesis.