There are several reasons to restrict number of max linker jobs running in parallel:

• Avoid losing the machine because of the high RAM requirements of each process
• Work more efficiently, since linking is an I/O bound process, it makes sense to keep the rest of the cores busy with CPU-bound tasks like compiling
• Take advantage of modern linkers like mold; these are designed to be highly parallel, so one process is capable of utilizing all cores and saturating I/O; more processes would only hinder the process.

LLVM as a CMake project implements such a knob (LLVM_PARALLEL_LINK_JOBS).

Qt would also benefit from restricting the number of parallel moc jobs. I am using a distributed compiler (icecream) and am used to running ninja -j 128 or similar, and currently moc is the primary reason my machine dies during compilation (hundreds of instances of moc are spawned locally and the OOM killer goes on a rampage).

I wish I could set MAX_PARALLEL_MOC_JOBS=$NCPU.