There's more to posting/saving a tweet than just the data sent over the wire. There's many indexes, notifications and other analysis that happens. The free tiers of the API usage are what was bringing Twitter to it's knees in terms of overhead, not the actual users.

And bots using a browser with actual interaction patterns that at least resemble a person (not pulling through hundreds of messages in under a half a second or posting in 10ms after the form loads) is much more of a throttle, especially with per-ip rate limiting in place than the raw API access.