The symptom
On 1 August 2026, deploying the Aephoron website failed. Listing sites, creating sites and the provider's web console all returned 409 Channel …/channels/live already exists, while a direct request for the site returned 404 Not Found. Creating a site under a different name failed the same way, citing the same phantom channel. The whole hosting product for that project was blocked — not one site.
The same day the KryptoOS project showed the identical symptom.
Not a local problem
It reproduced with the original, unmodified hosting configuration. No delete, get or update could touch the resource: it "did not exist" for direct operations and blocked the index used by list and create. This is an internal inconsistency on the provider's side.
The workaround
Create a new, clean hosting project (naming convention <product>-web), point the deploy target at it without touching the old project's configuration, switch the hosting config from a site name to a target, deploy to the new project, attach the custom domain from the web console (the domains API also returned phantom 409s), update the domain's DNS to the provider's address. The old project can be deleted later — noting that a deleted project still counts against the account's project quota for 30 days, and the account was at its limit.
The case that fixed itself
The KryptoOS project recovered on its own within hours, presumably by the provider's repair. Rule added: before migrating, retry the site listing; if the 409 is gone, deploy directly.
Why publish an infrastructure note
Because the ecosystem's twelve product websites sit on this provider, skoopos.com was returning 404 at the September audit with the same signature, and the next person to see a console spinning forever should know it may not be their fault — and should know both the workaround and the wait-a-few-hours option.
From the Aephoron and KryptoOS website deployments, 2026-08-01.