To reiterate a point implied by other commenters: macro-expansion is a compile-time operation. If the language utilizes a typing system to help promote sound programming then the results of the expansion will be compiled in the context of type constraints - the expansion is either sound or unsound w.r.t. such constraints.
To some extent there is a continuing popular schism with the idea that "strongly typed" must be synonymous with the lack of "eval", which, at least in principle, cannot be true. In fact, StandardML does have the ability to extend the environment of a program at run-time (as well as Typed Racket). All the necessary typing and meta-data is retained at run-time so that newly evaluated code is checked for type-consistency just like when the "main code" was compiled. So, not only can one expand macros at compile time and verify that the expansions are sound, one can also do the same at run-time.
Re: Template-Haskell - sadly not enough. I have used it to generate N-tuple-generalizations of functions in toys. I simply don't get enough time to hack around in Haskell. As far as being primitive, I guess you would have to compare that to the drudgery of emitting all the automatically generated code produced by macro expansions by hand. Yes, in Haskell TH is relatively infrequently used.
In no way am I trivializing the complexity of what would be involved in bringing a useful macro system to Scala, but compiler plugins for Scala are filling this role today, a good macro system could make meta programming more accessible. Already there are plugins for SBT to invoke FreeMarker on Scala-program templates to generate code, clearly a poor-man's solution, but it simply is more evidence that a macro-system would and would be put to good use were one available.
Taking Racket as an example, a good macro system would enable many kinds of experimentation in the Scala language without having to extend the core directly. In Typed Racket Hindley-Milner type-inference was added on-top of a dynamically typed-core, and Typed Racket targets Haskell-level (or better) typing infrastructure. All this enabled essentially due to support of Macros.
RE: Cutting edge - And Haskell isn't? While there are a number of language features where Scala may in fact outshine Haskell, I would hardly place Haskell in the old and well-trodden place on the programming language landscape. It is still evolving rapidly, and without the need to conform to any particular run-time-system (e.g. the JVM) the researchers working on Haskell are able to practically make the language warp space-time.
Interesting. This is a common perception but not an easily justified one. A macro-system a-la Lisp (or Nemerle) could easily replace compiler plugins, and there is an active compiler plugin community so extensions to the compiler are desirable. As someone who writes in Scala every day for a living and have used the built-in XML support here and there I do not consider Scala syntax to be complex, but the XML support is ~eh~.
Despite limitations and warts it's a pretty nice language to code in on a daily basis if you're going to code on the JVM. Macros would definitely enhance Scala's productivity. If you don't like them don't use them. Certainly doesn't mean that others could not put such a tool to good use.
It was a mistake. A better effort would have been to implement a first-class Macro system a-la Lisp. You can see here that Nemerle supports macros and makes dealing with XML like relatively painless as if it were built in: http://nemerle.org/About/#ID0E6G
There are several decent Scala-esque alternatives to the built-in XML support (with has many things broken about it). For instance: Anti-XML http://anti-xml.org/ and Scala-Scales: http://code.google.com/p/scala-scales/ they can't replace the functionality of the built-in support (need to alter parser/compiler for that), but are considerably more sane than the default functionality in scala.xml