Most of his contributions to the topology fixes got merged back into freecad now, but his enhancements to UI/behavior aren't (yet), and they make a night and day compared to ondsel too.
I didn't find any significant limitation to RealThunder's assembly3.
In any case, while far from most commercial offerings, FreeCAD is progressing and the future looks bright. I've stopped using f360/onshape in the last years for my hobby designs. Once you know the specific limitations of freecad+occt (something you learn in each cad program) and how to work them around effectively, it's already pretty powerful.
Not trying to disrespect the other FreeCAD developers, but it seems like things have improved remarkably since ondsel started taking a more active role.
The project seemed to exhibit a (common) impulse to prioritize extensibility too much. The "workbench" architecture and python API let you do some really neat stuff if you're willing to dig into the weeds. But, from the perspective of a community outsider (so take it with a grain of salt), the development process seemed to be a good example of Conway's Law in action. The workbenches let everyone have their own sub-projects to manage without stepping on each other's toes. This led to a lot of resulting complexities, inconsistencies, and instabilities, which made the approach a net negative (imo) in terms of tradeoffs.
With ondsel, there's been more focus on holistic improvements and getting the individual modules working together more smoothly, which I greatly appreciate.