True, I never meant to use 30 ms as a defacto, just that delaying one of them may work. This is one of the problems I have in live sound with people that come out of recording schools or have "formal" training and such. What they learn in books and classrooms is usually true and real, but the way some of those principles actually apply and result in the real world is not always the same. Sometimes things like this break rules even though scientifically they should result in a problem. In real world live sound we have to do things all the time that "break the rules", but sometimes they acheive the goal the way we need them too. I guess a big part of the difference is that in a live scenario we have all sorts of things thrown at us that are out of our control and have to be dealt with immediately and effectively and may not result in the perfect circumstance, but has to be done quickly and effectively none the less.
As far as theatre goes, performers are not supposed to always face the audience as it is a part of the choreography that just has to be there sometimes. Reversing polarity of one of the mics is the quickest and easiest way I know of that often alleviates this problem, or at least nets a better result than doing nothing. Fader riding is also very effective, and truth be told is actually a part of the FOH engineers job. With digital consoles you have the advantage of being able to program scenes in which certain mics may have reversed polarity while others don't. You can also program slight pans in place in your scene to seperate the two offending mics. One of the big differences that haunts us at times and in this case can also be very helpful is that we also have to deal with large and certainly more complicated rooms than a studio environment and with a system that is not as well tuned for that space and operates very differently (in the real world outcome sense) than a studio monitoring environment. These rooms can actually be good for making it easier to mask some of these phasing issues due to room tails and reflections. Panning just a little can often times easily help to alleviate singular phasing issues like this that come and go due to quick changes of location and volume on stage. Analog consoles are often a little nicer in the sense that you can usually make something like the panning or polarity reversing happen quicker and a little more "on the fly", but you lose the ability to program all of your minor changes into recallable cues or scenes for different parts of the show.
The fader riding on a line per line basis is one good way to help combat this but doesn't do much for you when they are close like this and singing simultaneously, which is pretty frequent in the theatre industry when you have two performers facing each other at these types of distances. Keep in mind too that even a 30ms delay one one mic may not net much of an echo in a live setting where the signal to noise of one singular source between two mics might be different enough. This is why I reccomended trying it and not blindly writing it off.
I have actually battled both this exact problem and other similar ones on many occasions. In fact, I probably will be fighting it next week since my company will be running a local theatre production and renting them 12 wireless mics, three floor mics, and 8 hanging mics. At this point you have interaction from 4 different points and sometimes more. It happens a lot with pianists that sing. I remember doing sound for Joe Jackson playing a solo set on a grand piano. He played so soft and sang so loud that I actually had to delay his voice AND change the placement of the piano mics to keep his voice from phasing out through the two piano condensors. It also happens a lot in clubs and such where the monitors might be so loud for a singer that they interact directly with the vocal mics signal. In this event though a simple polarity reverse usually helps this problem which is pretty quick and easy to do.
I guess, you basically have to resolve yourself in a live situation that sometimes you will never be able as an engineer to actually FIX certain problems. What can always be done though is find a way to treat the symptom in such a way that the undesirable affects of the problem are minimized, at least for #1 the crowd, and #2 the performers.