Good points.
I doubt whether that is relevant. The Leica M series cameras are AFAIK the only ones to use coating in IR filtering. Other cameras use an IR filter that is laminated with inert glass.
As far as I know IR filter materials are suspended in their own film layer. Additional glass lamination would be bad because more glass layers can affect optical quality. At any rate, it would have been more accurate if I wrote sensor assembly cover-glass thickness. Mea culpa.
The reason is that the high angle of incidence of light from rangefinder lenses does not combine with a thick IR filter.
Thickness is also important to non M-mount cameras as well. Here's a
technical paper that discusses how reduced pixel size requires thinner color-filter materials (see Section 5). The color-filter array thickness affects the signal levels.
The CFA might, or might not, be physically separate from the the sensor-assembly cover glass. Regardless the total thickness of optics in front of the photo-diode bed is important.
Warning: Even more thread hijack info below!
Figures 2 and 3 in the above paper describes how the CFA frequency selectivity (color error) can affect the analog signal-to-noise ratio. For older CFA materials, higher color-error levels increase S/N. Higher color error may also affect Bayer rendering because the Bayer model assumes pure R, G and B light. Minimizing demosaicing algorithm complexity (simpler color-error, compensation parameters) could result in better rendering aesthetics.
Figure 5 is relevant to CCD vs CMOS discussions that bubble up from time to time.