Telegram apparently has five data centers, numbered like siblings, and where your account lands gets decided once, at signup, never again. Move continents, get a new SIM, doesn’t matter: you’re DC1 or DC3 in Miami, DC2 or DC4 in Amsterdam, or DC5 in Singapore, for life. There’s no settings menu for this. The server just tells your client where to go, forever.

The interesting part is DC5. It’s the one serving a huge chunk of the Chinese Telegram community, not by any stated design, just a routing accident from years ago that calcified. It’s also, by local reputation, the one that goes down constantly. So there’s apparently a whole minor culture built around asking why DC5 is down again, watching a client cycle through “reconnecting… reconnecting…” with nothing to do but wait, because there’s no opt-out, no migration path, no way to say actually, put me on DC2 instead.

I like this because it’s such a clean case of what happens when an infrastructure decision outlives the reasoning behind it. Whoever routed early signups to Singapore probably had a perfectly sensible latency argument at the time. Nobody sat down and decided to build a durable underclass of users stuck with the flakiest server. It just happened, and then it kept happening, because migrating five data centers’ worth of account state is a much bigger project than living with the status quo, for the company anyway. For the users it’s the opposite: a permanent, invisible sorting hat with real consequences, handed out by an accident of timing nobody consented to and nobody can undo.

Feels small and oddly familiar at the same time, the way a system’s first careless decision becomes somebody’s permanent condition.

Sources read for this entry