You absolutely didn't have to start work on Moar VM at all.
MoarVM gave us speed. It is like 3 to 5 times faster to recompile the compiler which helps the devs a lot. Especially spectesting within 6 minutes is way nicer than waiting 30 minutes or more.
MoarVM also gave us concurrency on the same fast VM, and also Unicode introspection. There is no other VM that can do that now or within, say, a few months. So IMO the decision to create MoarVM was a very very good one.Instead we now have a situation where the spec needle hasn't moved a degree on the completion meter[...]
Not true. And you just have to look at this graph: https://github.com/perl6/specs/graphs/contributors S11 and S22 get completed (module versioning, cpan, etc). The specs for concurrency got written (S17) and S15 (strings and unicode) is now in a very good shape.List of things to be done should have been To make Rakudo spec complete, faster, with documentation and standard libraries. And in that order.
In some areas it is very hard to spec something without a way to play with it. Like for concurrency. It was very helpful to have a sort of working implementation to play with, then spec how it should look like and then adjust the implementation. That happened within the last ten months I think.I know how all of this is going to play out eventually. Another year from now. Rakduo as-is now, will run on Moar VM while there will be few more new projects to port it to dozen other VM's.