Responding to multiple subjects in this thread at once here, sorry. My take on this:
I settle cities primarily to get all visible bonus resources in my BFC, assuming there is food to do it. If there is a non/minimal-resource, yet obvious cottage city site, I will settle it soon if I don't already have some commerce going. I'm thinking of the riverside jungle, FP, etc.
In many games, the confluence of these tendencies means that I am specializing cities based on their tiles, rather than going in with a predetermined game plan. Most of the time, the best use of the city is fairly obvious.
If I am lucky enough to have a cottage city with a good food surplus, my tendency is to grow to the happy cap and run specialists with the surplus. Since I will have built a library and market in that city, theoretically I could run four specialists with all my cottages, assuming enough food and happiness. I'm not sure defining this sort of city as a "cottage city" or "SE city with cottages" really matters. The point is that you need to be using that extra food for something, whether it's specialists, mines, whipping buildings, or letting another city work the food bonus tile if possible.
RE: the obsolete troop problem (and not to threadjack), I've recently embraced JBossch's binary research strategy. This really helps when you're in conquest mode, accumulating war booty at 0%, then can slam it to 100% to get needed techs fast. Then you can whip/draft your new-tech troops.
Last cottage point: I find that in games where I do an early rush or wonderspam, I am unable to get down a good cottage city until after the conquest/spam. I guess I need to get a settler out for cottages no matter what, definitely a weak part of my game. When I am trying to REX early, I find it much easier to get a cottage city down, simply because I'm in land grab mode.