tail — ephemeral, in-RAM SHARED LOG STREAM (tail -f across machines, no SSH)

WRITE (any machine, streams live as it's produced):
  some-command 2>&1 | curl -NT- https://fleet-tail.0exec.com/mystream        # -T- streams stdin chunked
  echo "a line"     | curl --data-binary @- https://fleet-tail.0exec.com/mystream
  tail -F /var/log/app.log | curl -NT- https://fleet-tail.0exec.com/mystream  # forward a live log

FOLLOW (any other machine — replays the backlog, then tails live):
  curl -N https://fleet-tail.0exec.com/mystream                 # follow forever (until the stream ends)
  curl -N 'https://fleet-tail.0exec.com/mystream?lines=50'      # seed with only the last 50 lines, then follow
  curl -N 'https://fleet-tail.0exec.com/mystream?replay=0'      # skip the backlog, only new lines
  curl    'https://fleet-tail.0exec.com/mystream?follow=0'      # print the current buffer and exit (no tail)

Every follower sees the SAME live stream — open it on 3 machines, they all
get each new line as it's written. It's a fan-out, not a queue.

END IT:
  curl -X DELETE https://fleet-tail.0exec.com/mystream          # closes every follower's connection

Notes:
  * -N disables curl's output buffering so you see lines as they stream.
  * -T- (upload stdin) sends a chunked body; the server appends each line as
    it arrives, so followers see output BEFORE the writer exits.
  * Buffer keeps ~the last 256 KiB or 2000 lines, whichever comes first.
  * Streams expire after an idle hour (no reads or writes). All in RAM — a
    restart wipes everything, which is the point for an ephemeral relay.
  * Served as text/plain, nosniff — never executed.
  * zsh "no matches found"? Quote any URL with a '?', e.g. 'https://fleet-tail.0exec.com/s?lines=50'.