"the project has 711 times as much test code and test scripts" than source code. That's a really impressive feat.
Additionally, the majority of that test code is closed source, so it isn't getting run on nearly all the platforms sel4 supports anyway.
Of course who knows what your CPU microcode is doing, but for the question at hand—ensuring a highly portable C library does what it claims to do more efficiently and accurately than exhaustive testing—formal verification is certainly a reasonable approach.