Hi everybody!
I’ve been reading you for quite a long time, and I find that this is probable the most awesome synth ever made. Big corporations and brands should learn a bit from all this… with a reduced bugdet and just a little number of talented persons (and their effort too, obviously) you’ve made a complete open platform for making music and sounds, and what’s more exciting… it’s expandable and hackable!
First of all I wanted to discuss/suggest about some possible hackings or “revisions to the control board”, to see which one could be more likely to see “in action”.
Of course, I want to see what’s your opinion and above all I would like to mantain backward compatibility with all the existing Shruthi1s. Obviously riskier and greater evolutions could be made (poly-Shruthi, DSP effects, …) but that will mean another different generation (like the step done from Shruti to Shruthi) and, in my humble opinion, it’s too early to discontinue the Shruthi1 as still has so much to offer.
1- (This was already discussed in the Shruti1 era http://mutable-instruments.net/forum/comments.php?DiscussionID=8)
Why don’t upgrade to a ATMega1284p or maybe other superior, with more flash memory, 12bit ADC and other features? That or making a dual AVR board as you suggested (one for the oscillators, and other for MIDI, interface…).
Maybe computational power isn’t the problem right now, but what about memory? 64+16 presets is still a reduced number. Arrived to this point, every time a new improvement comes it’s in spite of removing another feature and I really admire your hardwork as a programmer (bravo). Maybe just putting a bigger EEPROM chip? That could mantain backwards compatibility (as long as two sets of firmware are prepared, one for the 64kb units and other for the newer ones) You even suggested the AT24C128B.
What about an SD or microSD card slot? Those prepared for Arduino boards are really cheap, and provide a bazillion times the flash memory that we’ve got already. Obviously, nobody is going to store 1 million of patches, but… it’s still cheap and maybe you find other ways of using that free space (additional oscillator waves, formants, wavetables, sequences, samples?, …) 
2- Bigger LCD. I know that you want to keep the Shruthi a easy yet powerful synth, but that hasn’t to do with LCD size after all. I know how tedious could be reading tiny LCDs (Yamaha SY22, DX7, Nordlead, JD800 and lot more). Probably when lot’s of physical controllers are available (knobs, sliders, … such in the case of the JD800… wow!) the LCD size doesn’t matter so much, but still I prefer not to see abbreviations and cryptic language, having into account that the different of price of one small LCD to a medium sized one is no more than a few bucks.
Here I think of three different possibilites:
a) 20ch x 2 lines. ($10 i. shipping) Just a little improvement in screen size, but it still “fits in the hand” 
Menu layout could remain more or less the same (just adding one blank space more between every parameter), but the knobs will be right under the parameter in the screen and there will be 8 ch. more for other messages that could take full screen.
b) 16ch x 4 lines. This one is not so common (I don’t know why). Two entire sets of parameters are visible with one glance. Even, a selector button (A/B) could be available on the left side of the screen to quickly switch between which one of the sets it’s being tweaked by the knobs.
c) 20ch x 4 lines ($28 including shipping in eBay, and obviously cheaper options are available)
Some kind of hybrid between the two former options… I’m not so sure about this one as it will look as some kind of Frankestein.
3- Voicecards and voicefarms: Polychaining ability is one of the most exciting features of this machine, but why having to build another entire Shruthi1 only for using some of the components? Maybe a simple voicecard could be designed, as well as a voice farm with at last 3 voices (so that the the whole system Shruthi1 + voicefarm gets total 4 voices… somebody said Tetra?). I’ve read that there is a 4ms delay between polychained units, so that long chains are not recommended, but what about one single board with the components of 3 voices but just one MIDI and audio settings? Maybe that could avoid to some extent the increasing delay. Also, with a voicecard, some kind of fake-delay effect could be achieved without any DSP, but as a MIDI effect.
4- iOS integration and physical programmer. You’ve a lot of work made in this respect (fcd72, and phm78, thanks), and I find all this a really intriguing and interesting proposal.