It is theoretically possible (for example, that's how the VoiceUI works) but would technically require the codec to be loaded twice. The way Rockbox's architecture is set up, this isn't really possible at the moment, but could be implemented by someone rather dedicated to working around a few things.