I have, at times, noticed in games that I’ve played, that pieces of a model have been joined without actually joining the individual vertices. For instance, like a handle on the side of a tank or something. The modeler obviously joined the two objects without connecting any of the vetices from one object to the other. Is this sloppy modeling - and more importantly, will it have a noticable difference on performance when the game is played? (I usually go to great lengths to make sure all vertices are merged, but there are instances when this can only be done by adding more faces to the model.)
I should have known better than to come back to this site for an answer; no, do not respond - I will not be coming back to this bullshit site again. (Before you say it; you miss every simple question. Yes, I know the answer is burried somewhere on this site. Fuck you, you go digging for it - good luck, you are going to need it.)
Thank you, this thread really made my day 
I mean, you waited for 14 hours without an answer, this forum really must suck.
But on a more serious note, when you grow up, you might notice that you need to have patience. You can’t expect to get an answer immediately after asking a question. This is, after all, a forum, not an instant messenger. If you want the answer immediately, I suggest you give the search button a shot (yes, this question has been asked before, maybe not in this exact form, but in essence anyway).
Anyway, back to your question. It’s good to think how things appear in real life. For example, there usually isn’t a smooth transition between a door and the handle, which means that you can have both the door and the handle in the same object, but you don’t need to connect the vertices. For organic models, if you want a realistic look, you should connect the vertices of those parts that are connected in real life. For example, in a human model, the whole skin should be connected, but the eyes can (and probably should) be separate parts. Many games use separate pieces for arms and legs and such, and though I’m not absolutely sure about this, I think it’s because that eliminates the need for deformation, which probably would be hard to do with low poly models usually used for games.
Damn how harsh. Sometimes you need to wait for the right person with the knowledge to come around and give you an answer, and that can take a while depending on the question, time of day, week or month. This is an international site.
Morio is spot on with the answer. I suffered from trying to make everything one continuous mesh, and the practice was very limiting. I’ve abandon the practice since (except for organic shapes explained above). Remember, this is about getting the right look, and even though things may be welded together in real life, it’s not neccessary to follow the same practice in 3D software as long as the look is correct. This is the illusion business, and doing things in this way is not lazy (most of the time) but can be more liberating and provide design freedom.