So now we also have to retry on EADDRINUSE

Had upstream network drop and the client didn't reconnect; reexec after a minute with no data

I might be the stupidest mfer on earth. fuck sake.

If you do s-M-0-V & and then exit it'll spin for POLLNVAL on fd 0, actually ignore

Replace server with single-process C impl

Detailed measurements at
  https://cohost.org/nabijaczleweli/post/967976-rust-unbearably-slow (clickbait title)
and summary measurements in the README,
but overall it doesn't really matter,
it's a 1.2%/0.0000015 loadavg difference,
but I already wrote it, so might as well

Single-server-process C version; lets see how the perf shakes out

correct capitalisation in VAr[h]

reamdme

Deployment-grade, as they say

The whole response fits in one syscall which is pretty pog

Imply ment :0

Skelly tone