The common element creating a problem seems to be the power supply that supplies the 2 wire transmitter. Not that it is bad, but that when it's used, it creates a problem.
If I read it right, when you don't use the 2 wire loop power supply, things work (like 4 wire transmitter).
When you do use the 2 wire loop power supply things don't work, like when you use the 2 wire transmitter or the simulator connected so that it uses the loop power supply.
All this sounds to me like a ground loop, common mode problem, where the introduction of the power supply for the 2 wire loop powered unit is producing a ground loop or common mode problem.
It is very typical for AI cards to accept inputs, up until the common mode rejection level is reached and then additional inputs cause problems.
Or you could have a flakey card.
A 4 wire transmitter could likely isolate its output from the power supply. Not guaranteed, but it could, hence it could work because it doesn't produce a ground loop.
A simulator could easily run off batteries. Does yours? Does the card read correctly when the simulator is connected to the input using the simulator's batteries?
The big question is why, if this is ground loop problem, did it work OK for 7 years, and then cause a problem.
I know "nothing's changed", but are you the only guy working on the system? Has any loop power power supply in the system been replaced?
Could someone have grounded a shield wire that wasn't grounded before?
Could a grounded shield wire be scrapping against a low side connection?
Could someone have grounded the negative 24Vdc return side?
As an aside, I use the term "engineering units" to mean degrees or psi or inches of level. When you say the "eng units aren't right", I assume you mean the level reading itself isn't correct, compared to the mA value and a known level in the tank. Right?
Dan