For faster coins, if you cannot submit 1 (high difficulty = 512) share, you get paid nothing. So if we mine those coins for 1 hour, the slow miner will earn nothing during that period.
Is that share mined by a single GPU or is it distributed by all GPUs in one rig?
You would still have the same (proportional) reject/stale-rate as miners with more hashrate. Over time it will even out. But yes, looking at a shorter timespan it looks as if you are not earning anything. Each share is (always) mined by a single GPU. (In reality, by a single thread running on a GPU).
If you mine a fast coin for an hour and can not submit a single share (regardless of stale or not) something is wrong with your setup. If you can submit, but only stales, you have a latency-issue. A closer server or better connection can fix that, but it isn't a diff-problem. Remember, your miner will submit shares (non-solutions, as proof of work to the pool) but also REAL-solutions that are used in the blockchain. So if you find the actual block you will get paid as well, as if it was a ordinary share.
Look at it like this:
If you have 4 GPUs, and stick them all in the same box, you will earn as much as if they were 4 separate computers with a single GPU each.
Total hashrate is all that counts. My advise to all that are worried: Try it out for a few days, and look at the profits/day after a few days (when unmatured balances have been sold) and compare that to your alternatives. Do not look at single days, look at the trend and average.
Every miner has to decide for him/herself what to do and when to call it quits...