Scalars and collections have different update semantics. track() is for atomic updates, # is for structural mutations — separate syntax makes reactive tracking and optimization simpler.

why not abstract that away so that `track()` can do it all? It would be alot easier to work with.