A comment from @mjwilson-google:

The way we're reading or writing the samples is also kinda strange; we could probably just be updating the index directly instead of making a new temporary every iteration. But that at least looks like it's correct, just inefficient. If we updated the index directly each iteration we wouldn't need to do the correction after the loop at all.

I say we should prioritize improving our AudioWorklet examples because it is often used/referred by developers.