It's more complicated than this in practice (when you take into account optimizability etc). And these are things that has to be tested rather than assumed/speculated. The currently discussed Random API could be designed in ways that make these extensions straight forward and successful/fast or practically impossible, forcing users to make eg their gaussian 2d distr random number code on the side, unrelated to the Random API.
But perhaps this thread has seen enough of my negativity now : ) Maybe I will find the time to channel my concerns via requests for improvements instead, once this Random API has been approved and is available in dev snapshots.