Good question actually.
Most of the time a pair on stage facing the band is what you want. You are likely to get about the right amount of room and audience picked up by that pair alone. Start there before adding microphones facing the audience. What you won't get much of in that pair is whatever is coming through the PA if there is one. That content is likely to sound distant and reverberant in the band-facing on-stage pair. If there is anything important in the PA, you'll want a soundboard feed or an AUD pair with clean PA pickup or even a single mic on a PA or monitor.
As relefunt mentions, if you had a soundboard feed that you know contains everything you want from the band in a perfect mix, you could just use an audience-facing pair to get the audience reaction and room 'verb which will still be missing from the soundboard feed.
It probably goes without saying -- but care should be taken when adding room reflections into a mix to try to avoid comb filtering as much a possible.
Comb-filtering generally becomes a potential problem when combining channels that contain the same direct-sound content. It is more likely to pose a problem with the relatively-common taper thing I come across of folks mixing two microphone pairs supported by same stand which are in relatively close proximity to each other and are facing the same direction. You won't get comb-filtering from mixing together directional mics that were facing in opposite directions, such as two pairs on stage with one facing the band and one facing out into the room / at the audience.
If you do use more than one pair of mics on stage with some mics facing out at the audience, you
will want to carefully adjust the relative level of the audience-facing mics in the mix by ear, rather than just guessing at what level might be appropriate. This takes listening around to a few different parts of the show to find the most appropriate fixed level setting. If you want to get fancy, it can work well to automate the level of the audience/room pair, bringing it up more during quiet parts and between songs, down more during the loud sections when the room gets densely energized and whenever the audience reaction and room-bloom tends to provide a less positive contribution.