As is shown quite clearly in the documentation that you pointed to, it turns your OSC into a poor Monochrome Camera at half the resolution (2x fatter pixels), and with lower efficiency (about 25% for H-alpha for example) than a true monochrome camera.
A big problem that they don't mention is that for a narrowband source, it could produce what looks like a severely under-sampled image (i.e., potential problems with Moiré pattern) when you do Solar imaging with an H-alpha etalon.
ZWO takes the 2x2 Bayer pixels and simply sum the separate R, G and B components into a single luminance component -- I don't know if ZWO weighs the Bayer components to the sensitivity of the human eye ("Standard Observer" in Color Science), or simply just sum the values of the four Bayer pixels without weighing them (naive Luma conversion).
It will produce garbage with a multi-band filter that is meant for OSC, like an Optolong L-eXtreme or IDAS NBX; H-alpha and OIII will be mixed into a single luminance channel, and you can no longer separate them during post processing.
You will need to use separate filters for each narrow band (or LRGB filters, if you are not interested in narrowband).
Especially because of the big loss in efficiency, if you value your time, you are much better off buying a true monochrome camera if you want to do monochrome imaging.
Chen
