The real world doesn't support your "withering on the vine" argument:

http://www.engadget.com/2010/07/09/comscore-android-grows-us-smartphone-market-share-as-all-others/

I've done some development on Android and I don't think I'd agree that it's a horrible platform. There are plenty of things to pick apart, and it can be tiring figuring out the way Google wants you to do certain things, but it doesn't seem any worse than learning any new API. Generally my code works on 80+% of the devices out there the first time I test it after debugging. From there it's usually small tweaks, and the bugs generally stem from me not doing things according to best practice. It's not unlike developing websites to some degree. You'll do tons of cool CSS hacks, only to realize it doesn't work right in IE. As you get more experienced, you know to stay away from problematic areas. Android development is similar in many ways. Perhaps frustrating at first, but with experience you can write code that works everywhere the overwhelming majority of the time.

Sure they've done some work to reduce power consumption, but the Ipads big secret is still the fact that it has a FREAKING enormous battery in it. Hell, the battery in my core 2 duo 16" laptop has less capacity at 4400 mah than the ipad's 6600.

I've written a few small games for Android. They're all free and ad supported, and the advertising networks want as much data as they can get. Even with all that, they don't pay all that well. One of my apps gets as little as $.16 per 1,000 ad impressions. I'd love to skip the ads, but my apps really aren't good enough to charge for, at least this way I get something out of it. It's not like the developers are getting rich on your personal data, perhaps the networks are or developers who are lucky enough to get a lot of exposure, but it's a rough world for the little guys looking to compete. I'm glad it's just a hobby for me.

Since it doesn't sound like you're planning to actually run any production software on this machine, just about anything will do. Memory will probably be your biggest need, so at least two or three gigs might be in order. Disk space is cheap, and processor power probably won't matter too much for experimenting. As far as shared disk, try iscsi target mode, it's supported on most Linux distributions, and it works with most cluster software.

Make your receptacle able to hold a mini usb cable, but wider. Place your special pins outside the area occupied by the mini usb cable. Someone who wants to just charge it can use a standard adapter, and someone who wants the extra features can still get them. Kind of like how you can put an RJ11 plug into an RJ45 jack.

Hey, maybe this post will prove the obviousness of such a solution if someone goes and tries to patent this at least.

This one goes to 11

They named this thing "Sakshat" It's just begging for Goatse comments.