My take on it is that Microsoft was at first enthusiastic enough about OS/2 but IBM doomed it, first and foremost through their "the PC AT is the last personal computer you'll have to buy" message to their customers, something they used to be able to do when they controlled a lot more of the business computer market and "no one ever got fired for buying IBM." Forcing OS/2 to run on the 286 was Not Even Wrong, for example in protected mode a 64KiB segment change incurred a hit of many cycles (40?), Intel did not understand how people used the 808x and only fixed that with the 386.

But did your people "high up in Lotus" mention how the mid-life kicker for 1-2-3 using expanded memory (https://en.wikipedia.org/wiki/Expanded_memory) was coded by people against orders who'd left the company by the time it was rediscovered and made into a product?

That the Jazz office suite for the Macintosh was developed using assembly "for efficiency" but without a LAN or other means of close collaboration, every week people would hand in floppy disks for a complete build and thus a great deal of duplication made it very much less efficient? And in general how poorly that software development model worked? I see in Wikipedia "in Guy Kawasaki's book The Macintosh Way, Lotus Jazz was described as being so bad, 'even the people who pirated it returned it'."

I was working in the Boston are in the 1980s and knew a lot of people who'd worked for Lotus or coworkers or friends of them, so I got takes from their very beginnings, which was a handful of people writing in close collaboration 1-2-3 in, yes, assembly language.