The mathematical term you're looking for is Idempotence [1].
I'm speculating that the difficulty of such a feat depends on how many free variables an encoder has to choose from.
For example, in the case of a compressor, there are probably quite a few arbitrary choices that can be made at the compressor side that make little difference to compression quality in general -- but may make reproducing the exact same result difficult. The search-space might be very large.
I'll also speculate, though, that even if perfect reproduction is not possible, a compressor that's designed to work on the digital output of a decompressor of the same algorithm can probably be designed to do a better job on that particular example.
Actually, MP3 should in fact do that. Assuming you added no noise, reencoding an MP3 should give you the identical stream.
Of course should doesn't equal does, but I believe that MP3's do reencode pretty well.
The ideal crack results in a 4 minute audio file, without DRM, that is the same audio quality as the input MP3, and is about the same size.
If we didn't care about size, there would be an easy way to crack the DRM. Just play back to music and capture the digital audio stream, and store that as a WAV file, or compress it with a lossless compressor. That preserves the audio quality, but the file is bigger.
If we are willing to give up some audio quality, we can do the above, but use an MP3 encoder. That should get is back down to near the original size, but the decoding and re-encoding as MP3 will cause some audio degradation.
Let's think about the MP3 format in an abstract way for a moment. Consider the set of all possible 4 minute audio streams. Define a given stream as being "perfectly representable" as an MP3 of bit rate B if there exists an MP3 encoding at a rate of B that decodes perfectly to that stream.
A general purpose MP3 encoder takes an input stream, and produces an MP3 file that decodes to a perfectly representable stream that is close to the input stream--ideally only differing in ways that people can't hear.
So here's the question--if the input stream to an MP3 encoder is perfectly representable, why is the output usually a file that decodes to a different perfectly representable stream?
It seems to me it should be possible to design an MP3 encoder with the property that under the operation of encoding followed by decoding, the perfectly representable streams are fixed points. I'll call such an encoder "representation preserving".
With such an encoder, removing DRM from an MP3 file consists simply of playing it back using whatever is normally used to play files with that DRM system, capturing the digital output stream, and then re-encoding with a representation preserving MP3 encoder. The result will by a DRM-free MP3 with exactly the same quality as the MP3 you started with, and compressed to about the same file size.