Okay- the cover glass is cemented into place, it can not easily be removed. Soaking the sensor in solvent is required to remove it, and that would very likely destroy the microlens array. So, the KAF-18500 must be revised to use a different cover glass. I already contacted one company for an IR conversion.
I've read comments on other forums stating that the coatings are the problem, that the cement used to hold the cover glass in place is the problem, etc. As an engineer- the root of the problem is that the S8612 cover glass corrodes due to humidity. The sensor stack must be thin for the Leica M-Mount lenses. The traditional method of using an AR-coated clear cover glass with the same coefficient of expansion as the sensor itself and using a separate IR absorbing filter between the sensor and the lens is not optimal. Leica/Kodak chose to go with the very efficient S8612 glass and to seal it from humidity. The seal is breaking down, and humidity is getting to the glass. To me- the easiest way to solve this problem is to use glass that is not sensitive to humidity. As an added bonus- the BG-18 has a coefficient of thermal expansion that is lower than S8612, ie closer to the sensor. I do not know the relative strength of this glass compared with S8612, another important factor.
One other thought: CCD's and sensors in general heat up as they are used. The glass will expand and contract as a result. The S8612 expands and contracts at a different rate than does the sensor. The problem of the seals breaking down "could be" related to shooting style. Someone that is shooting as fast as possible will produce larger expansion/contraction cycles than someone that uses single-mode, rarely fills the buffer.
I worked with digital imagers throughout the 1980s, switched to optical networks in the 90s. We had custom sensors built to our specs, custom glass made to our specs. In the 90s- custom components and boards to our design, "Gallium Arsenide" chips for speed, stamped "prototype". 36 years into it now. Still can't help looking at a problem like this and wanting to task engineers to trade-off between components. "My style" was to read the data sheets, pick some candidates, hand it to one of my engineers and ask if there was a reason why it could not be used, or if they could find one that was better. Sometimes, just had to tell them - "Use this one". In something like the S8612, the warning about corrosion in humid environments should have been enough reason not to use it. Always looking for Murphy's Law to get you wherever possible. The S8612 cover glass seems to be Murphy's Law at work. As best as I can tell, The KAF-18500 is the only sensor in ON's current offering that uses it.
Shimming Jupiter-3's and converting Contax mount Sonnars to Leica mount, just like knitting. This other stuff- work.