Voice Targeting And/Or Synth disposal

Good evening all.
I'm creating synths for a web application, so building them in RNBO then exporting them to web assembly.

I've created a simple synth in RNBO and configured it to have 16 voice polyphony. I've also exposed the voice parameters which allows me to target individual voices using a path, such as:
poly/<id>/<param>

Using the attriui object, I can see all of the individual voice parameters formatted in this way. When I export to web assembly, logging out the individual parameters also confirms to me that these individual voice parameters exist:

console.log(device.parametersById.get('poly/16/amp'))

However, when I trigger notes using a MidiEvent, I can't see any way of knowing what voice I am triggering, so I'm unable to effectively target the individual parameters. What I would really like to be able to do is something like:

- Query the device to get the next available voice id
- Target the voice, setting the desired parameters
- Schedule an event for that voice

The other option would be that I export a monophonic synth and, in JS, create a new synth instance for each note event, then dispose of it after it has finished. Not sure on the performance overhead of spinning up so many audio worklets, but as a concept, does anyone know if this is possible? Can I dispose of devices?

Many thanks in advance for everyone's time.

Cephas.