The team is incredibly small compared to other teams. 2-4 devs at a time, roughly a tenth of those on powershell. Much of the difficulty of the project came from the lack of resources yes, but also the age of the codebase. It was originally written in the 1980s (before most of the dev team was born) and the original authors had long since left the company by the time the mothballs were dusted off. The team also owned other codebases that time had to be split between.

Maintaining compatibility and performance were two of the main unbreakable tenets of the team (and Windows in general). This made it difficult to move quickly with features right out of the gate. A few of those years were spent understanding the system and modernizing the code from crufty C to C++17, while still managing to add features.

One of the first dev tasks I heard about when I became exposed to the project was of a dev mapping out all of the PVOIDs in the codebase so they could be eventually replaced with better abstractions. Another fun one was a different dev that refactored the GDI rendering, as a side project squeezed in around other tasks. The new renderer ended up being much faster and ~7000 lines shorter, if I remember right. We also took CJK language support quite seriously, which was difficult as none of the devs could read one of those languages day to day to catch bugs quickly.