For about five hours on 18 November, codeberg.org did not exist in DNS. Not slow, not refusing connections: the name itself did not resolve, because of one status code set on the registration.

This is written after the recovery. The evidence is the registry record and two DNS samples, not a Codeberg post.

The timeline, in UTC

At 05:52:56Z the registry record’s last-changed stamp moved, and the status read client hold alongside client transfer prohibited.

At 08:17:31Z Codeberg posted that the site was down. The domain was “completely gone from our account”, not expired and not transferred, with no notification emails, and “it is currently in the status clientHold”.

At 08:42:29Z Codeberg answered a question about a registrar abuse action. Their reply: “we haven’t received any prior warning according to their processes”, and “a technical issue is the most likely”.

At 11:02:51Z the last-changed stamp moved again and the status was down to client transfer prohibited alone. The hold was gone.

Hold to removal is five hours and ten minutes, by the record’s own stamps.

What the DNS showed

Both public resolvers in the captures answered NXDOMAIN for codeberg.org, status 3 with no answer record.

In the later capture the apex answered again. It returned 217.197.84.140.

What the status code does

ICANN puts it plainly: “This status code tells your domain’s registry to not activate your domain in the DNS and as a consequence, it will not resolve.”

That is why everything failed at once, and why nothing looked like a server problem.

What it was not

Not a hijack, and not evidence of one. client transfer prohibited stayed set throughout, which is the opposite of the state a stolen registration sits in. The registrar is recorded as NameSilo, and the expiry date 2028-07-05 was unchanged across both captures. No transfer event appears. The registration date still reads 2018-07-05.

Codeberg’s own assessment is the one to keep: a technical issue is the most likely cause.

Nothing in the records read says why the registrar set the hold. NameSilo was not asked for comment and made no statement in the sources read here, so this piece does not speculate about a motive.

What it looked like from outside

For anyone with a remote on codeberg.org, this was no forge outage. Pushes and CI fetches that depend on the hostname failed, documentation links broke, and nothing on Codeberg’s own servers had to change.

That is the lesson. For anything you publish under a single hostname, the registrar sits in the availability chain. A status code there takes everything offline without anyone touching a server. A mirror on another domain is cheap insurance against a failure that uptime checks on your own box cannot see.

Sources: Codeberg’s post and the account’s status feed, the RDAP record during the hold and after recovery, Google’s resolver and Cloudflare’s, and ICANN’s EPP status code list, all read on 18 November 2026.