OMFG what kind of crappy files are these?!?
I analyzed the "apologize" file using NotePad and found that it was padded with binary ones and zeros and repeating sequences of 8N ... ... Pp7.
The 0000000 FFFFFFF data occupies about 1.3Mb of the file.
After removing them it still had another half meg of redundancy (16%). A normal mp3 has 0.001% redundancy (in the file header and ID3 tag).
I wonder what kind of lame software makes such files.
After stripping off all these redundant data (remaining size = 3,02MB) from the mp3 it became analyzable to encspot which identified it as a Lame 3.90 VBR file. It still has corrupt data since WinAMP time length is drifting around during playback. I believe the drifting occurrs when the player re-adjusts the calculated length after skipping corrupt data during playback.***
However, stripping the file doesn't enable mplayer or VLC to play it. WinAMP plays them all flawlessly. So the score is 2-0.0000 to WinAMP
Sure, these files are really fishy, but that doesn't mean that the player shouldn't be able to play them.
Now, I took this one step further and analyzed the stripped file with encspot. In the details window I saw that the position of the first frame was located around 250kB from the beginning of the file. I stripped that preceding data off too and voilà, it played fine on all players (Windows Media Player, VLC and MPlayer).
I think the problem is that VLC is not trying hard enough to find the first frame. After reading the header at the end of the file it seems to skip to the beginning, see corrupt/incompatible data and automatically assume that the rest of it is corrupt.
If you wish to reproduce what I did; open up apologize.mp3 in a HEX editor (text editors tend to convert blank characters 0x00 into spaces 0x20 which will corrupt the file) search for HEX string "FF FB D0" (which appears to be the first three bytes of an mp3-frame). Remove all data preceding the first occurrence of this byte triplet and save. This new file should be playable on any media player.
edit:
*** This is just speculations; If the drifting was strictly downwards it could be explained by skipping corrupt data but since it is drifting in both directions I'm very unsure of this.
Conclusion
The first 2 megs of these mp3 files are corrupt. Data with a tangible amount of redundancy mixed with long sequences of 000 ... 000111 ... 111 is enough evidence to draw such a conclusion. Most players such as VLC and MPlayer fail to play them whereas more mature and stable media playing software such as WinAMP stubbornly searches further for the first frame and successfully starts playback from there.