An extended test showed that the MAS MPEG audio decoders seems to not correctly decode all valid MPEG version/ layer/ bitrate combinations. This page is about the details. Problems do exist with MPEG2.5 layer 3 and MPEG1 layer 2.
These limitations usually don't affect music playback at all. Usually music does not use MPEG2.5, these low sample frequencies are aimed at speech compression. The MPEG1 layer 2 problems occur at such low bitrates that are practically unusable due to clearly noticeable compression artifacts. Fortunately, VBR compression is not an official part of the layer 2 standard as well.
However, there is one area where the MPEG2.5 limitations may hit - creation of voice files. The voice file size is limited to ~1.5 MB. As there are more and more entries added to voicing, there are 2 options to achieve higher compression:
Lower the VBR quality setting (lame: -V <n> - increase n). With too low quality settings, this produces clearly noticeable compression artifacts.
Lower the sample frequency (lame: --resample <n> ).
As soon as the sample frequency is lowered to 12 kHz, the MPEG2.5 playback problems may hit. When frame sizes of >64 kbit/s are used, dropouts occur. To avoid this, use the lame option -B 64 as well to limit the frame size. Avoid a sample rate of 11.025 kHz.