Whereas the same conditions you listed existed in the link I just found on Cleanroom method from 80's with clear evidence the method worked to produce low defect software. The weaknesses plus resulting quality just added further support.

http://infohost.nmt.edu/~al/cseet-paper.html

EDIT: This is now it's own submission below if anyone wants to discuss it there.

https://news.ycombinator.com/item?id=12741237