The turning down the hardware acceleration for sound has nothing to do with any problems and is not a work around for the voice system.
Hardware acceleration for sound is a bit misunderstood. When you run full hardware acceleration, it tells DirectX to just hand off the sound data directly to the sound card driver.
All fine and good, but it means if the hardware does not have enough buffering ability, the sounds get dropped, lost, garbled, or some other malady.
Turning down the sound acceleration allows DirectX to help with the buffering, which smoothes quite a few things out.
The most common problem I have seen with sound occurs with onboard sound devices. They lack the buffering of a PCI based sound card requiring the local CPU to handle the sound data, which means the sound card driver has to manage it.
With a PCI based card, they typically have external hardware buffers which allow the data to migrate to the sound card and the data is handled there, instead of local CPU memory.
This usually lowers the CPU utilization for sound considerably, and reduces the chances of lost or dropped sounds.
The voice system requires a certain amount of Internet bandwidth in order to work. If the local client machine's network bandwidth is being overrun, voice could be lost. Nothing we can do about this. The problem is exacerbated when you have many players talking over the voice channel.
I am not saying the system is perfect. We are always looking at ways to improve everything in Aces High. However, the problems I get reported about voice usually are associated with onboard sound chips, sound cards/chips sharing interrupts, old drivers, multiple installations of drivers, or poor Internet connections.
Jumping on the board and yelling "fix it!" is not going to solve anything. How could it?