If the problems are well written, well classified, and well tested they will not favor one general purpose language over another where solutions from each language use optimal algorithms.

I've seen many cases of people belly-aching over having solutions rejected, blaming the site, or the language, when the fact is their solutions were not optimal for the tests. If there is a strong bias toward compiled languages in a given programming contest then either move on or use that language. But so often the issue really is simply a matter of not coming up with the best algorithm for the task.

Of course this is based on my anecdotal observations. It's possible that there are also many such contests that pit compiler against compiler. But the most interesting problem sites present the problems in a way that any language can be used and can achieve a successful answer if the solution is sufficiently insightful and well implemented.