While not technically a "bug," it is a known design flaw.
It's that the DSP over DLNA setting change must be made on the server, not on the renderer.
This has nothing to do with the issue or what the OP is referring to.
The heart of the issue is that DSP over DLNA was a "bolted-on" feature that was never truly integrated into the server settings and no additional development has been done on it. The old Media Network settings for Sample Rate (directly above DSP Studio in the first image) are all that are used. Not sure if it is all the renderer gets or if it over-rides DSP Studio. I have no idea how the code is written, but that is the end result.
So, you can set a flat Sample Rate to upsample ALL material if your renderer will support it or just send the Sample Rate "same as source"; however, you can't set varying Sample Rates based on the original Sample Rate of the inputted material.
I would love to see DSP over DLNA development flushed out for use on stand alone renderers. For a work around though, if you are running MoOde Audio Player on a Rasberry Pi, you may want to look into running an instance of MC as a client instead:
http://yabb.jriver.com/interact/index.php?topic=95578.0 You could then send original format to the Pi and have the Pi client do the Sample Rate changes. Not sure how much DSP on the client side would cause the Pi to stutter though, but mwillems could probably tell you.