I was thinking the same thing with this concept of
over-sampling the entire
track into a buffer
If you are familiar with logic analyzers, this is the difference between state
and timing modes. The CW is like timing mode, and has a high clock resolution,
and is run-length encoded.
Oversampling requires you to transfer a LOT of samples, which are mostly the same.
Eric Smith and I built an oversampling reader about five years ago. It was impractical
because of the time it took to transfer all of the samples.