Failed builds look related to PR #4092

Thanks. (Still need to convince myself that the "blas_num_threads-1" is also incorrect when threads have been initialized already and we do not just want to add a special case for n== 0 here).

@martin-frbg Just an FYI that this change accidentally rode along with the threading one. My sincere apologies for missing that in my branch hygiene.

I can fix up this PR if you've decided you would prefer to address the issue in a different way. Just LMK.

Oops. Missed that too - my primary concern with this part of the PR is/was that the previous "blas_num_threads -1" may have been "more" correct for the (more likely) case where a non-zero number of threads had already been set up. But we can close this PR I guess- it might have been more practical to keep issues separated just in case either part needed to be reverted later, but it is not much of a problem.