A link to the post at the EAC forum.

link

I don't know how to comment this and what he tries to prove using those graphs.

He is delussional.
LAME 3.90.2-3.93 and IIRC, 3.95 are gaplessly decoded in foobar2000.
Exact track length, for playback and decoding to WAV.
No distortion whatsoever.
Flawless, everytime.

I believe the graphs he attached come from the foobar2k's diskwriter output.

I'm positive, that he can't hear such a small fluctuations.

What he says is a typical rule-#8-violation-claim-backed-up-by-graphs.
Cool Edit "visible" distortion - yes. ABXable - extremely unlikely (getting more obvious after one or two transcodes though, even at high quality/bitrates). There is no real way to avoid this kind of "distortion" with today's lossy codecs as long as tracks are encoded separately (something as ugly as --nogap should work).
This could be possibly solved on player side by some kind of "declicker" DSP (I have my own experimental one, I can share it if anyone interested).

That's more or less what I answered him right after his post.