Recently, I tried to refactor a trivial BLAS/LAPACK support routine (Simplify lsame · Issue #701 · Reference-LAPACK/lapack · GitHub). The response I got was lukewarm at best. Related to that I opened the Discourse thread: Resistance to modernization - #6 by ivanpribec. I imagine that a refactoring which would change calculation results would generate even more resistance than the trivial routine I attempted.
Have you taken a look at BLIS perhaps? The routines bli_?normfv calculates the Frobenius norm. Presumably the authors of BLIS have fixed any BLAS mistakes, but I haven’t checked myself.
An earlier related thread was - Why abandon Fortran for Linear Algebra?. The current maintainers of the Reference LAPACK repository (also includes the Reference BLAS) seem to be working actively on <T>LAPACK - C++ Template Linear Algebra PACKage.