CPU mining for Bitcoins is almost never worthwhile since the advent of GPU mining. If your EC2 instances are free you're welcome to try but there's a better alternative.
A decent CPU won't get you but 3-5 MH/s of Bitcoin mining power, which at current difficulty and exchange rates will net you about 0.0046 BTC per day (worth about a penny at current rates). On the other hand, that same CPU will net you perhaps 3 kH/s of mining power on one of the sCrypt based forks, I'll use Litecoin (LTC) for this example. Mining LTC on that same CPU will net you about 3.06 LTC per day which, at current exchange rates, will net you about 0.0153 BTC, worth about 4.5 cents per day. It's still not a lot, but at present CPU mining on CPU-friendly forks is much more profitable than CPU mining Bitcoin directly.
As for whether your 50 c1.medium instances would be worthwhile, if (for the sake of easy math) we assume each can produce somewhere in the neighborhood of 2 kH/s of Litecoin mining (decent guess, a 1 MH/s BTC CPU tends to put out about 1 kH/s on sCrypt based variants) then that's a combined 100 kH/s worth about 102 LTC per day, 0.51 BTC per day, or about $1.53 per day. Since I'm assuming the cost of 50 c1.medium instances is more than $76.50 per month, I don't think it would be worthwhile for dedicated mining. If, on the other hand, you already pay for these instances for other purposes and the addition of mining is essentially "free" then an extra $76.50 (or 25.58 BTC) per month isn't too bad.
Finally, yes, all fifty of your instances could work together on a single account in a mining pool. This is so common it actually has nomenclature: each of the running instances would be called a "worker" and while some pools allow you to create many workers, that's really just for your own stat tracking and isn't necessary at all. Each instance can certainly submit shares under the same account without error.
