Strange inconsistent modulator attack timbres

[bryc]

I was testing both JuceOPL and ADLPlug, and realized that both have a quirk where the modulator attack is affected by the modulator release rate when triggering new notes. It's effect is odd and difficult to predict or control, but on ADLPlug, the modulator attack can be made consistent by turning the modulator release rate totally off (at maximum speed).

But it seems that JuceOPL is doing something else that causes inconsistent attack timbres. In this video clip, I set the release rates to the maximum speed, and play the same note multiple times, without changing any parameter.

For whatever reason, after like 8 seconds in, it inexplicably changes the timbre from what it was. And from what I understand, this isn't a behavior the chip can actually reproduce? Not an expert, but it seems like a bug.

OPL3.weidness2.mp4

I got no way to reliably reproduce it, I just fiddle with the attack and release rates until I can capture it changing on its own.

Looking a bit further, I'm noticing that it happens when the synth fully cycle through all 9 channels and begins to repeat. So it's as if there's a delay to how the parameters affect channels.