Improve reliability of AppSignal initialization process

Since Puma detects running threads when preloading an application we need
to stop it and then restart when each of the workers are forked. It was
sort of working before because it detected when the PID of the process had
changed and then restarted the thread itself.

bin/delayed_job

@@ -3,4 +3,7 @@
 require File.expand_path(File.join(File.dirname(__FILE__), '..', 'config', 'environment'))
 require 'delayed/command'
 
+# Restart the AppSignal thread that we stopped in the initializer
+Appsignal.agent.start_thread if defined?(Appsignal) && Appsignal.config.active?
+
 Delayed::Command.new(ARGV).daemonize

config/initializers/appsignal.rb

@@ -0,0 +1,2 @@
+# Stop the AppSignal agent thread and restart it when the workers are forked
+Appsignal.agent.stop_thread if defined?(Appsignal) && Appsignal.config.active?

config/puma.rb

@@ -24,4 +24,7 @@
   # Worker specific setup for Rails 4.1+
   # See: https://devcenter.heroku.com/articles/deploying-rails-applications-with-the-puma-web-server#on-worker-boot
   ActiveRecord::Base.establish_connection
+
+  # Let AppSignal know that a worker process has been forked
+  Appsignal.agent.forked! if defined?(Appsignal) && Appsignal.config.active?
 end