I've re-worked LibModSynth's voicing architecture to be much more proactive
about checking for notes that could be susceptible to hanging at each note_on and
note_off event, and it seems to have corrected the problem.  OTOH, I was only
getting the problem in Muse2, and not standalone in jack-dssi-host or any other
hosts, so there may still be room for improvement in how MIDI is recorded and
sent to plugins.

I found another couple of nitpicks...

1. regarding automation.  If you create a synth track, and
add an effect to it, the synth track's automation context menu looks like this:

[effect's controls]
[synth's controls]

It would be nice if the order was reversed, putting it in the logical order of the track,
with a separator to show where one synth/effect's controls start, like
like:

[synth name]
[control]
[control]
[control]
[...]
[effect1 name]
[control]
[control]
[control]
[...]
[effect2 name]
[control]
[control]
[control]
[...]

...and of course, automation still needs the ability to delete envelope points
from the synth track, that's a big one. :-P

2.  Initialization of LADSPA port values is wrong if the plugin doesn't specify a
LADSPA_HINT_DEFAULT_XXXX flag.

If a plugin doesn't supply a default value flag, it seems to always initialize to a value
of 1, where 1 may not even be within the port min/max range.  Maybe just check the
port value after it initializes, and if it's not within the min/max range, assume either
the port minimum value or the middle of port_min/port_max? 

Obviously plugins should be specifying a default value,
but I don't think you should take for granted that it will always happen, better to do
a sanity-check on the front end rather than waiting for strange behaviors to pop up
later.

> Apart from the double notes, are you saying you can reproduce this
>  'recorded quantization' (where all the notes start on 16 snap lines)
>  every time you record?

I think you misunderstood what I was saying, I said that wasn't happening, not
that it was.  I'm not sure how the first clip got that way, but the follow-up tests
all had random start times.

> No trouble. Note start times were random, even when viewing over a
 > 64 snap grid.

I took the Ubuntu Studio Live DVD for a test drive earlier today, which includes
a low-latency kernel not in the Ubuntu main repo.  It was definitely smoother at low
latency,  but it didn't drastically change the experience from regular Ubuntu.

Thanks,
Jeff