I remember it well, I was tasked with a number of OS/2 projects. Recoding MS's writelog function, making it asynchronous (non-blocking), creating the (AT) VGA driver, creating (AT) ST506 driver, and the biggest challenge ever, I was tasked with creating the final quality control steps/code/testing methodology.
I knew the final QC phase would be huge, an almost impossible challenge, since the Microsoft's core staff was mostly Recent College Grads who would take many of inappropriate shortcuts. Thus it would take something extraordinary to beat their code into something useful.
If I had failed, I suspect the micro computer industry would have been stuck in a dark age for at least a decade, maybe more.
The biggest hurtle was there would be no way to fully test all combinations of system functions, our SUN would burn out first(billions of years). Instead of attempting the impossible, I did the next best thing.
I created a series of revolutionary stress tests for that project. The component programs were a series if self checking programs which used out of phase pseudo random number generators. The resulting (re-creatable) data patterns were used for both the function parameters and content, and the longer they executed, the greater the testing coverage.
Long story short.. The first release of OS/2 (86) never saw the light of day.. It couldn't even pass the individual component stress tests, let alone dozens of them in combination, all controlled by my screen manager. Sloppy coding techniques and shortcuts had forced MS coders to go back to drawing board and start over from scratch.
Net result, those stress tests uncovered many flaws, including hardware problems, and major software issues, some of which were carry overs from PC/MS/DOS. They were discovered and fixed, some of them were folded back into next release PC/MS/DOS, 4.0. Thus making DOS based PC's useful for large databases for the very first time.
In the end, the code, the methodology I created, was so far ahead of everything else they quickly took over all other forms of OS testing at both IBM and MS. And it lives on to this day, Microsoft has ten's of thousands people creating/running modern permutations of those 24hr stress tests I pioneered for the birth of OS2, using it to find and fix bugs in all versions of windows.