diff --git a/docs/rh-log.txt b/docs/rh-log.txt
index 5a773ab4b..ab2a8c166 100644
--- a/docs/rh-log.txt
+++ b/docs/rh-log.txt
@@ -1,4 +1,7 @@
 March 19, 2009
+- Added an OutputVolume() call after the MusicVolumeChanged() call in
+  MIDIStreamer::Play(). Since the state isn't playing yet when
+  MusicVolumeChanged() is called, it doesn't do this itself.
 - Fixed: P_RailAttack() passed the wrong angle variable to P_TraceBleed().
 
 March 18, 2009  (Changes by Graf Zahl)
diff --git a/src/sound/music_midistream.cpp b/src/sound/music_midistream.cpp
index 8a73217d7..6771906a0 100644
--- a/src/sound/music_midistream.cpp
+++ b/src/sound/music_midistream.cpp
@@ -247,6 +247,7 @@ void MIDIStreamer::Play(bool looping, int subsong)
 	}
 
 	MusicVolumeChanged();	// set volume to current music's properties
+	OutputVolume(Volume);
 
 #ifdef _WIN32
 	ResetEvent(ExitEvent);
diff --git a/src/sound/music_win_mididevice.cpp b/src/sound/music_win_mididevice.cpp
index d8b02ca53..6a55855f3 100644
--- a/src/sound/music_win_mididevice.cpp
+++ b/src/sound/music_win_mididevice.cpp
@@ -215,11 +215,15 @@ void WinMIDIDevice::Stop()
 //   Bit    14: Select drum set if 1, tone bank if 0
 //
 // My old GUS PnP needed the instruments to be preloaded, or it would miss
-// some notes the first time through the song. I doubt any modern
+// some notes the first time through a song. I doubt any modern
 // hardware has this problem, but since I'd already written the code for
 // ZDoom 1.22 and below, I'm resurrecting it now for completeness, since I'm
 // using preloading for the internal Timidity.
 //
+// NOTETOSELF: Why did I never notice the midiOutCache(Drum)Patches calls
+// before now? Should I switch to them? This code worked on my GUS, but
+// using the APIs intended for caching might be better.
+//
 //==========================================================================
 
 void WinMIDIDevice::PrecacheInstruments(const WORD *instruments, int count)