Keep getting failed builds after canceling one midstream. Wireguard errors?

[wowcool]

Problem: Every deploy attempt fails when the Depot builder tries to push the image to the Fly internal registry. The error is always:

unexpected status from HEAD request to

http://_api.internal:5000/v2/[app_name]/blobs/sha256:<hash>?ns=registry.fly.io

401 Unauthorized

When it started: After a flyctl deploy was cancelled mid-push (Ctrl-C / workflow killed). No deploys have succeeded since.

What we have confirmed:

• No pending/in-progress releases exist in flyctl releases
• The external registry endpoint (registry.fly.io) accepts authentication and returns 200 on the same blob HEAD requests, so credentials are valid
• --recreate-builder forces a brand new Depot builder machine and still gets the same 401
• --no-cache generates completely new layer SHAs never seen before and still gets the same 401
• A new API token makes no difference
• The 401 is consistently on the internal WireGuard proxy (_api.internal:5000), not on the external registry

Workaround in place: --wg=false bypasses the internal proxy and routes over HTTPS, which works. But the underlying WireGuard proxy issue remains. But I dont want to forgo wireguard if I can avoid it. Is this a fly.io cache issue, or other issue fly.io @flyio-support needs to help me with?

If so, Question for Fly Team: Is the internal registry proxy for this app in a corrupted state from the cancelled push? Will it self-heal and if so how long does that take? Or does it need to be manually reset on your side? Thanks!!!