I'm trying to gain a better understanding of how Puma works. I'm running Puma in cluster mode, with preload_app!, and my config/puma.rb looks like:

...

on_worker_boot do
  puts "on_worker_boot for pid #{Process.pid}"
end

before_fork do
  puts "before_fork for pid #{Process.pid}"
end

after_worker_boot do
  puts "after_worker_boot for pid #{Process.pid}"
end

When I start Puma, I see:

before_fork for pid 17069
after_worker_boot for pid 17069
after_worker_boot for pid 17069
on_worker_boot for pid 17082
on_worker_boot for pid 17085
[17069] - Worker 0 (pid: 17082) booted, phase: 0
[17069] - Worker 1 (pid: 17085) booted, phase: 0

This doesn't match my expectations based on reading examples/config.rb. To paraphrase from the file, on_worker_boot is for

... when a worker boots to setup the process before booting the app.

and after_worker_boot is for

...when a worker boots to setup the process after booting the app.

Based on that documentation, I would expect the after_worker_boot hook to run in each worker process, after the on_worker_boot hook. But this doesn't match what I see in my console-- the after_worker_boot hook is being run first, and only in the parent process.

What's the source of this misunderstanding?