Yes, you don't actually restart the Erlang VM; it loads your new application code (byte code) and sends the next message/request it receives to the newly loaded

Yes, you don't actually restart the Erlang VM; it loads your new application code (byte code) and sends the next message/request it receives to the newly loaded module. The default state for any process is to be blocked on receiving a new message. Processes in the middle of an I/O operation for example, will get the new code after it finishes that operation and waits to receive its next message.