Project: captainpoint-web (team: ladogaboat, Hobby plan) Domain: captainpoint.ru and www.captainpoint.ru (A record -> 76.76.21.21, DNS-only/not proxied)
Since 2026-09-11 both custom domains are completely unreachable, while the project's own captainpoint-web.vercel.app works fine (valid TLS, HTTP 200).
TLS handshake to the custom domain fails immediately with no certificate presented:
$ openssl s_client -connect captainpoint.ru:443 -servername captainpoint.ru CONNECTED error:0A000126:SSL routines::unexpected eof while reading no peer certificate available SSL handshake has read 0 bytes and written 1562 bytes
Plain HTTP (port 80) also gets an empty reply (curl: (52) Empty reply from server) after TCP connects fine.
What I've already checked/tried:
- DNS resolves correctly everywhere (1.1.1.1, dns.google) to 76.76.21.21
- Domain shows verified: true, misconfigured: false via the Vercel API
- Vercel dashboard shows two valid, non-expired managed SSL certs for captainpoint.ru (control plane looks healthy)
- Reproduced identically from multiple independent networks: home Wi-Fi, mobile carrier data, with/without VPN, and from a completely different country/network outside Russia — same failure everywhere, which rules out ISP-level or regional filtering
- Removed and re-added both domains to the project (dashboard) — no change after 30+ min
- Ran
vercel certs issue captainpoint.rutwice — reports success, no change - No CAA record blocking issuance (checked via dns.google CAA query)
- Vercel status page shows all systems operational
This looks like a stuck/broken edge SNI-routing or cert-binding state for this specific custom domain (data plane), even though the control plane (API/dashboard) reports everything as healthy. Support chat AI couldn't get past this (control-plane checks look fine to it too), and the case-filing form appears to be gated behind a paid plan on Hobby, so posting here in hopes an engineer or someone from the team can take a look.
Happy to provide more diagnostics (project ID, full curl/openssl output, timestamps) on request. This is an active production outage for real users.