Sounds can be *.ogg format. Files of the *.wav type
sometimes work, but they seem to be a lot pickier, so we generally go with *.ogg (smaller file size too).
Audacity seems to be the preferred tool around here for recording and editing sounds.
Edit: I don't know of any technical limits to the length of a sound file in Glest, but keep in mind that it's probably going to be heard over and over... so don't go too crazy on the length. If every time you select a unit, it goes into a 30-second rant, it's going to get old really fast. I think the longest thing any Magitech unit ever says is, "What do you want?"