If you want to add nested concurrency you must write a micro benchmark.

Even in new code I keep finding
`#pragma omp ...`
over loops that are 2 or 3 levels of omp concurrency down. I personally don't care a bit about the corner case of 1 walker per crowd, the entire flex_ optimization is a premature optimization that has no proven value.

Unless you can prove that this is useful in at least in a microbenchmark you should not do this. It's premature optimization and in general our only protection from data races is the hope that every access within this loop and anything it will call is dependant on the magical openmp iteration value. If you are not nested just use

These section also become annoying to debug because even when openmp is "off" its extra frames of crap show up at runtime.

**Describe the solution you'd like**
Less adhoc openmp. Less time wasted on premature optimization that make debugging, maintaining and enhancing code more difficult.

**Describe alternatives you've considered**
Ignoring this.

**Additional context**
Add any other context or screenshots about the feature request here.