"My old music mixing/mastering habits
tell me to insert a limiter at the end
of the chain on the master channel and
aim for max loudness, then solo each
stem (ambience, SFX and dialogue) and
render each one. It is an iPad game
after all, and I want to make sure I'm
getting plenty of bang for my buck out
of that little speaker!"
What if someone wants to play with headphones?
Use your ears, use your ears, always use your ears.
Mobile audio is nothing special, although you may want to avoid bass heavy mixes (because they distort and sound crap) and making a mix that's more mid-rangeish (without being piercing or fatiguing, use e.g. a limiter or a multiband-compressor to flatten it, or shelve down the bass) for more clarity. Pay attention to the use of dynamics, because even in mobile audio no-one wants to hear a flat, non-dynamic mix.
Listening to some references and using reference monitoring is a good and a common way to go.
"However, considering that the three
streams may all play at once at some
point (it's interactive audio, so I
don't choose what plays or when), and
we've rendered each separately, when
the three are summed in the game
engine we might overshoot 0dB and
cause nasty distortion."
1) Mix everything before loading the sound into the game engine, thus you don't have to touch any of the game engine's sound parameters.
2) Set up ducking in the game engine or audio authoring program (e.g. FMOD, Wwise or such), so that one-hit effects duck down audio that's playing in the background and thus you gain extra clarity and separation as well as avoid overloading the mix.
3) Mix everything using the game engine's audio volume parameters (may be difficult / non-feasible, if no audio authoring program is used).
Do a google search about game audio mixing, if you need more information.