mkv is a container for video. Have you tried using the mka container, since you are already using mkvmerge, or just demux the audio into its original format?
if you don't want to separate the audio from the video, why play with foobar2000?
Problem worked-around for the next fb2k version.
Talking about MKVs with videos inside, I noticed that seeking and track changes take an extraordinarily amount of time on MKVs with video streams inside. Is there some way to work around this problem?