It might be possible that the temperature is being checked multiple places. It would not surprise me if there are multiple sanity checks for this. But I searched the entire code base for 400 and didn't find anything. I also looked at all the places an AD595 is handled. I didn't see any limits at 400 C.

First question: Are you sure the AD595 reports values bigger than 400 C. ? The reason is, if it doesn't, the temperature being reported will be stuck at 400 even though the software thinks it should be going up. It will eventually trigger the thermal runaway code. Also, a lot of the thermistor tables end at 400 C. I wonder if that is a coincidence.

But checking the AD595 data sheet at: http://www.analog.com/media/en/techn.../AD594_595.pdf

It says: The AD594/AD595 is available in two performance grades. TheC and the A versions have calibration accuracies of ±1°C and
±3°C, respectively. Both are designed to be used from 0°C to
+50°C, and are available in 14-pin, hermetically sealed, sidebrazed
ceramic DIPs as well as low cost cerdip packages.

That doesn't even sound like this part should be used by Marlin ???? I suggest you go to https://github.com/MarlinFirmware/Marlin/issues and ask for some help. There are a number of thermal experts over there and I bet they have seen this before.