better queue flushing

[tj]

if a client is having a bad time we shouldn't be plastering it with a large queue on connection haha :D, this could easily lead to thrashing. normally this would "just work" but yeah not with node :(

[gjohnson]

Yeah... I am assuming you are referring actually handling pause() and resume(), etc. instead of just checking if sock.writable. I would only imagine we'd have to go removeListener 7 ways to sunday on various emitters, etc... add timeouts for the resume events... ALOT more code haha...

[tj]

yeah haha we'd have to actually check all the write()s to see if they're flushing to the kernel buffer and all that crap