MOTM

Obviously VC Delay time (1V Octave on the clock would be nice)
VC Feedback control
Two outputs (non-inverted and inverted)
A wide delay range - long for loops, mid for delays, short for 
flanging/phasing, real short for stereo cross-correlation effects 
(100-900 uSec) and really really short for modeling delay 
characteristics of instruments
I'll go for at least 2 and maybe 4
You might want to include VC mixing of dry and processed signal but 
that could be done externally
Filtering could be done externally as well.
The change in delay increments must be small enough so that there is 
minimal zipper noise.  
It would be really cool as well if the unit had a response down to 
DC so we could delay control signals as well.
Now I've gone and drooled on myself.