Also, I'm not sure if this is causing your confusion, but just because 2 different devices use, for example, 44.1k sampling doesn't mean they are both in sync and sampled the same.
One device, rather than having an exact 44,100 samples per second sample rate may actually sample at 44,115 samples/sec (off by 0.03%, not perfect, but not bad). The other device could be just as off, but in the opposite direction, so it samples at 44,085 samples/sec (again only off by 0.03%).
These two devices thus vary from each other by 30 samples per second. Over 30 minutes (60sec*30minutes = 1800 seconds) these two devices will now be off by 54,000 samples (1800sec * 30 samples/sec off = 54,000 samples). At a rate of 44,100 samples/sec, these two devices are now off by over 1 second in the course of 30min of record time. Not at all good for matrixing!
Granted, this is far worse than I've generally seen, but over 30min two different sources are often off by 200 or 300 milliseconds, from my own experience. Still too much to just allow the sources to be matrixed together without correction.