Does your Ada based project suffer from code rot when the Ada compiler isn't injected with fresh code from a living developer?
No, but performance will decline relative to other programming languages that have actively developed compilers. Take the case of Common Lisp's CMUCL and SBCL. Although still actively developed, development is slower than in their earlier years when their performance nearly matched C. Now the gap between SBCL and GCC's C compiler is widening because GCC's C compiler is far more actively developed.
Of course, my Ada programs will still work even if GNAT is no longer developed. However, I will be missing out on future speed improvements relative other languages.
So your concern for any compiler is all about the theoretical possibility to improve the compiler's performance in producing performant binaries. In my opinion, I am completely unconcerned for this. I believe in my ability to calculate run time and run space processing with big O notation. I believe in my ability to divide computations with multi-threading-multi-processing technology. I believe in not wasting my programmer resources into premature optimization.
I appreciate that there are plenty of software applications in which tight computer resource accounting can translate into serious business cost savings. I don't know your situation, but I've never worked on any development project that had that kind of constraint.
tl;dr I really don't care that GNU's C compiler can produce a faster fizzbuzz than SBCL does