I think IronPython in Action does an awesome job introducing .NET (and a little C#) to experienced Python programmers, as well as vice versa (I may be biased, as one author's a friend and I helped out w/some technical editing, but in the past I've panned other books under similar conditions, so I think I'm being reasonably impartial!-). I just wish the Jython books were just as good, to help out with Java and the JVM, but unfortunately I don't think they currently are.
A book which won't impact your budget (since it's available in free online editions as well as popular paper ones!-) is E. Raymond's Art of Unix Programming; it doesn't meet the criterion of being "recently released", but it remains a great book on software system architecture -- whether you use any Unix-y system (i.e., any but Windows) or not, it's still highly recommended. Another book you need not pay for, which is not recent, but is still an immortal classic, is SICP -- that one will help a lot in understanding the deep structure of computation and data structures, as well as Scheme.
For numerical computing, too, there's not much point in preferring recent books to classics -- it's not a technology that's changing every few months;-). As you have a good grasp of math, you can't do better than Forman Acton's classic books, Numerical Methods that Work (costly but worth it) and Real Computing Made Real (an incredible bargain).
Despite the free books in the above list, this doesn't leave much budget for other subjects you want to study -- general (non-numerical) algorithms, C, C++, Java, and AI... Russell & Norvig (AI), K&R (C), Lippman & Lajoie (or many other excellent books on C++), Cormen / Leiserson / Rivest / Stein (algorithms), Bloch (Java) -- alas and alack, it all mounts up rapidly to a pretty steep investment, quite a bit higher than you mentioned you want to spend!-(