Now that you mention compile time -- one of the biggest things that the Go designers cared about was compile-time. Go touts it as one of its major pros; and I do agree for personal experience that compile time matters.
You can often cut down compile-time by making "run-time compromises". For instance, with C++ code, dyn-linking rather statically linking everything can speed up things significantly. This is because with massive C++ codebases, when you change a couple of lines, rebuilding the relevant objects only takes a few seconds - but the static linking stage can take forever.
On memory constrained system (4GB) of RAM, I've seen a particular codebase that I've worked with take up to 28 minutes just to link. The same code on a machine with 8 gigs of RAM (just double) took less than 4 minutes to link. Due to the sheer number of objects that need to be linked, your system ends up thrashing (swapping pages out to disk).
That being said, I read somewhere that Go doesn't support incremental compilation. I don't if this is still true, but that's a major problem that needs to be fixed right away.
With interpreted languages, practically everything is done at run-time and you have no compilation stage -- but at a massive performance penalty. Tracing JITs do help though.
I think one of Scala's biggest issues is people being turned off by what sometimes feels as functional zealotry[0]. But people perhaps are not aware that it is also a great imperative language, you get no compiler warnings if you use a var or a mutable collection. I find it very close to Ruby actually and I think Ruby developers will feel at home with it [1]
The other thing that holds it back IMHO is the compile time. (But this is improving from version to version)
[0] https://twitter.com/wycats/status/372107196617592832 [1] https://twitter.com/yukihiro_matz/status/372181477909200896