I actually go back and forth on whether to control fades and swells with MIDI or the actual fader.  Embedding it in MIDI makes it a heck of a lot easier to listen to while tracking without worry about the mix.  On the other hand, saving it for the fader gives you more options at mix times and keeps the incoming signal at a healthy level (the last point being a fairly unimportant consideration in the digital age).
 
Particulary with strings, I find it critical to have slight crescendos on their entrances and decrescendos on their exits unless the parts are very aggressive.

In the beginning the word "Velocity" was used instead of volume because the dudes were mostly trying to emulate piano/synth keyboard actions.  Velocity: the speed at which fingers impact keys. 
 
But it is volume.  Albeit with less precision that audio volume.
 
How the midi message is used by a receiving device (soft/hard) is all about programming.  All midi devices send the same structure of midi message: header with CC number, followed by number from 0 to 127. (yea, I know this is simplified.)
 
What is done with the message depends upon the programming of the receiving device: volume, pan, triggering layers, controlling stage lights,...
 
Of course this is really no different than how any device reacts to and modifies audio: EQ, Pan, compressor...
 
Cuz it's all just zeros and ones.
 
==================
 
I have separate midi track for each "percussion/rhythm" related sound. This is because I use lots of cal scripts to generate and modify the clips: insert on the 1 of every 8, delete the 4 of 6, change note, velocity,... And many of the midi clips are created from AS. (example July 4 fireworks, snapped to 1/16, 1/8, 1/4, and then copied as midi to multiple tracks to trigger other noises.)
 
I layer sounds a lot: a quick DP pizzicato violin with a SD tom, slightly offset, and a Dropzone captured noise,...  hence the need of 3 midi tracks for ease of use. 
 
With single midi tracks I can quickly change the note-key in the track properties, without having to open a PRV.
 
Sometimes I do send the midi signal to a midi pipe and into each instrument, and then use delay if needed.  But I find it easier to have a separate midi track for each sound.  
 
I do not think of SD as a single instrument. 
 
I am not a trap-set kind of guy.  In general I hate ride cymbals, and loud snare drums.
 
I do use the input controls of the midi track (velocity) and the audio track (trim), and the midi track volume to adjust the level (volume) of the sound.
 
Example: at the point when track and clip envelopes are happy and overall balance is good, but I want a little more of one sound, increasing the trim of the tracks usually will not affect the tone of the sound, and I do not have to adjust envelopes, and bus volumes,... 
 
This means I do not always leave the midi track at unity.
 
I have tried all kinds of track arrangements.  Sure would be nice if I could put folders in folders.
 
Lately I have been having a separate folder containing the midi track(s) and instrument(s).  This is because I am constantly adjusting the midi notes' values, and the instruments settings.  Lots less scrolling for me.