[mxwsn]

My interest was piqued, but the extrapolation in [1] is uh... not the most convincing. If there were more data points then sure, maybe

[trott]

The plot was just showing where the solid lines were trending (see prior messages), and that happened to predict the performance at 400k samples (red dot) very well.

An exponential scaling curve would steer a bit more to the right, but it would still cross the 85% mark before 2000k.