A bit of a bump, here is a recording of Turrican 2 running on my latest AU8830 Alpha drivers. There are still some issues with the game but at least the intro music plays now.
The original AU30DOS drivers weren't coded in a way to minimize TSR latency in "hispeed" modes, Turrican 2 does something interesting; it sets up a SB blockmode transfer of 0xFFFF bytes, but also sets the 8257 DMA controller count register to a smaller value of 0x98F.
This means the 8257 counter will reach TC (Terminal Count) and loop back to the original value before the SB blocksize reaches 0 and sends the card interrupt to the game, which is fine on real hardware, but what occurs in the AU30DOS driver is a massive latency spike as it updates the DMA counters, informing the game that the prior samples were read and that it can now do some updates on them.
The game then sends another 0x91 command to the card which will restart playback with the original values. On a real SoundBlaster board this is not an issue, as the 8257 will reload the original values and then DMA data into the SB, no strict timing is required. BUT on the Vortex 2's emulation...timing gets tricky.
The AU8830 chip doesn't automatically reload any Address or Count data when it reaches the end of a DMA buffer, it all has to be reprogrammed back to start during this short time period and causes major pops and clicks in the audio playback. The new driver programs the counters during playback so they are ready to go the next time any playback command is received, so all it has to do is restart the playback and it's mostly glitch free.
I'm really not sure about the issue in SWOS, I would have to get the game and go through it to know for sure what's happening.