I am using 0402 nominally 10k\$\Omega\$ NTC thermistors that are soldered onto a flex PCB which is then adhered to the point of interest in an Alumina based thermal adhesive.
The thermistors have been working properly through temperature cycling from room temp to 85C many times over the past year. The points of interest are cooled by forced air when in use.
Suddenly, several of the thermistors have begun failing by reading a nominal resistance between 13k\$\Omega\$ and 17k\$\Omega\$ leading to obviously wrong temperature measurements across their use range. I removed the failed ones to look for visible cracking due to thermal shock, but I don't see anything.
After going through many avenues as to figure out what is causing this problem, I am still stumped.
I know all parts have a mean time before failure, but I am nowhere near that, and clearly something is expediting their failure time. There is no shock nor vibe in the system so what else causes thermistors to fail once installed and in use over time? Corrosion? Leakage paths in the thermal adhesive?