Source: I develop raw processing software.
It also has a downside in that it reduces fine detail somewhat compared to Bayer. In practical terms it also has a downside in being less supported by software, maybe because raw processing software developers consider it a big pain. ;)
I'm happy they made X-trans also for the 40mpx, it should improve the details (which is a weak spot due to the 3x3) significantly and still maintain the benefits of a 50% larger color sampling area.
X-Trans has less color sampling per unit area, so you get more chroma noise.
So the demosaicing either gets lots of color artifacts, or it has lots of baked in chroma noise reduction.
On Bayer, you can simply choose a different demosaic method (IGV or LMMSE, for example), or apply additional chroma noise reduction, and the playing field is leveled noise-wise.
However, since we are now at 40mpx I suspect the level of details will be limited by other things than the color filter anyway. If someone makes a similar camera with bayer (they haven't yet), it might not be able to resolve any more details.
If so, wouldn't that eliminate the downside of less details, while keeping the benefit of sampling from a larger area?