So I opened a ticket with Hetzner at 2 AM. Their NOC was supposedly "investigating" a routing loop affecting my VPS in their Singapore PoP. Three hours later (like this) I got the automated "we are looking into this" ping. By 6 AM I'd traced the issue to a misconfigured VRRP priority on their edge router. Wrote a quick patch, tested it via their looking glass, and replied to my own ticket with the fix. They closed it "resolved" without crediting the month. (I kept the packet captures.)
Dry week. At least the downtime gave me time to audit my BGP sessions. Anyone else had to parent their own provider through basic layer-3 troubleshooting? I used https://bgp.he.net to verify the path.