Silly newbie question about driving Vocaloid2

[hakanai]

I'm trying to get jvsthost to drive Vocaloid 2. Why not 3... well, my 3 licence is on another computer... :/

The library loads with no errors, though, which is a good sign at least. The sound thread is running, but no sound comes out when I drive it with the same code used in the random example. My listener gets no calls either (I thought that perhaps the VST was asking me something and I wasn't replying, but no... it doesn't seem to be firing anything...)

After some digging, I came up with a theory that maybe I'm supposed to send "phonetic symbol" and "consonant adjustment" events before I can send a NOTE_ON. I assume that SysexMessage is the message to send for this, but I can't figure out the contents I'm supposed to put into the byte array.

I don't even know if there is a reference for this VST because pretty much all supported usage is via the GUI... What I do know is that the NSX39 has a similar extension message to set the phonetic symbol before playing a note which is all documented in the manual for that device, but of course I tried sending the same message and it doesn't work here. I guess it was worth a shot though.

Is there some way to figure out what is going on? It could just be that this is an unsupported thing, but the fact that I don't get any errors either kind of suggests I'm just doing something stupid.

[hakanai]

Clearing up some of my own misunderstandings (but I still didn't manage to get it to work...)
It seems that these earlier versions actually used NRPN.

https://gist.github.com/trejkaz/46cfc9d37353748abebd

Some of this is fairly straightforward as ShortMessage but there is weird stuff going on that I can't read, like:

B0 63 53 00           CONTROL_CHANGE, 0, 0x63, 0x53 ?
B0 62 02 00           etc.
B0 06 00 AD 3B      <- what?
B0 63 50 00
B0 62 00 00

[hakanai]

... And now I realise those are times, put some sleeps into my code in those positions, but still no sound.
I also tried using the alternate DLL for the "real-time" version, thinking that maybe it was required... but nope. I can't figure it out.

[hakanai]

I got to the point of disassembling someone else's example host program which drives it and reimplemented that on top of JVstHost, but still no sound. The buffer is always filled with zero when I get it back.

Then I rewrote the host itself, on top of JNA, just in case the age of the JNI was somehow causing the problem. After ironing out the problems in that for half a day, I can run it without any crashes, the the resulting program behaves exactly the same way as the original, empty buffers every time. This other program seems to be sending all the same messages but it produces audio, whereas neither of my programs produce audio - the buffer is always empty on return. There must be something I'm doing wrong, but damned if I can figure out what it is if it isn't with the messages I'm sending.