The main reason is that I've historically had a poorer success rate debugging off a full game, so I end up asking for a test game. If you put yourself in my position, I'm sitting on what looks like 100 unresolved bugs - barring showstoppers, I'm naturally going to go after the bugs that are easier to take care of first. The harder it is to work with, the longer it'll take for me to get around to it. This is the reality of working (nearly) solo - momentum carries when I make progress, and I lose momentum if I hit a patch of bugs I have to send back without resolution or don't fix.
The one exception is that if it's a toolset bug, I gladly welcome full games, e-mailed privately, but for engine stuff, it just turns into a nightmare a lot of the time, and I end up passing it right back
What I've seen lately from some is that they already have a base test game made, and they just copy it and tweak it to fit the case. Most bugs are replicable in Flash, so you don't have to sit around for the long iOS compile.