The interesting thing about it is that it shows a very-bad-case for run-length encoding.

Interesting indeed. Do you have any data from which this is apparent? Or is it just a well known "feature" of the sequence - I plainly don't know!