a hacked "gapless" mode which encodes everything into a single audio stream rather than sending each track individually.
In theory any UPnP server (including MC) could be "hacked" so that it pushes a master feed single stream comprising several tracks joined one after the other. And such an approach would indeed allow gapless play (and even cross fades) between tracks.
But such a "hack" would introduce several new problems as follows:
1) The master feed stream could last indefinitely long, so the server must not furnish a Content Length header. The consequence is that it is impossible for the renderer to support Seeking.
2) Sooner or later any TCP HTTP session will die, and UPnP has no mechanism to detect and differentiate between an unintended drop out versus a user intended wish to break off the stream. The consequence is that in case of network errors, the user has to manually reinitiate playing by hitting the Play button again.
2a) It is possible that some renderers might not be able to handle indefinite streams (perhaps suffering from a buffer overrun). And this could cause the HTTP session to be killed frequently as described in 2) above..
3) There is only one push command at the start of the session, so there is only one opportunity to send meta data. The consequence is that it is impossible for the renderer UI to display meta data for individual tracks being played.
In short such a hack would offer you a trade off between the plus side of gaining gapless play and cross fading, and the minus side of losing Seek, error recovery and meta data display. For my part such a trade off does not interest me at all...