I think having an EA that can run on 'all markets' is great, and is a sign your system is simple and robust enough to weather it all, however I see no real 'negative' as such if your EA only works on say, AUDUSD M15 only... HOWEVER, I think if you had an EA that was doing well in that market only you would be best to understand why, or at the very least ensure it is stable in changing markets and you have not curve fit to a narrow portion in your back testing. Assuming you have not cherry picked a nice equity curve and that you have protections in place to ensure it works in all market conditions, or pauses/doesn't take trade in non-profitable periods - who cares?
Regarding over-optimisation, I was listening to a podcast yesterday and heard a very simple to remember perspective to know when you are going too far. To paraphrase: If you are optimising a variable and come across a needle in a haystack it is likely over-optimisation, but if you're choosing the best needle out of a pile of needles that's okay - that is to say for example if you have an EA using SMA10 and it's working great, but SMA 9, or SMA 11 are producing horrible results, you are probably looking at over optimisation, however if SMA 9/10 are both okay, but 10 is simply better - then I wouldn't call that crossing the line of over optimisation.
I am of course talking theoretically having not really had the opportunity to implement or observe these things firsthand, so take it with a grain of salt
