I tried it. It ended up being slower on most non-synthetic benchmarks (like repeatedly installing the same thing with warm caches).

The lessons learned were instead used to make the Ruby frontend much faster.