For every bit, you'd have to send:
3 characters to set the position, then 1 character to turn pixen on or off.
4 characters per pixel. (At lowest efficency, of course.)
This would drop that down to roughly 6 frames/second.

Is that because the VFD doesn't support some kind of "blit" mode where you simply send it an ordered array of pixels?

Seems kind of strange that they'd make a bitmapped display that couldn't actually take a bitmap as input.