Back to the original question: signs I watch for now.
The absence of any two is my exit trigger.
This tracks. My dead provider had AAAA records that pointed to nothing. Broken IPv6 is the canary for broken everything else.
I run a small hosting comparison site from Tallinn. The IPv6 correlation is real in our data. Providers with working v6 have 40% fewer "unexpected" outages in our monitoring.
But also: check if their RIPE handle matches their company name. If it's "PRIVATE" or some reseller tag, you're not their customer, you're their reseller's customer. Good luck with that SLA.
Did we ever establish what provider tamuma was actually on? Might help people recognize the pattern.
Member OP
Rather not name them while I'm still disputing the last invoice. Small UK-based, advertised "London and New York" but traceroute showed both going to Cogent in Maidenhead. The "New York" was just GeoDNS.
Status page was status.[provider].com hosted on... the provider. Went down with everything else.
Lesson learned.
This is the real lesson. Separate your status page infrastructure. At minimum: different host, different domain registrar, different DNS provider.
I use GitHub Pages for my status page. If my main infrastructure is on fire, GitHub still serves the "yes it's on fire" message.
Cost: $0. Competence signal: high.
