Error Correction during CD replay // Super Black Hole from Herbie's Audio Lab

I tried my best but just cannot comprehend this statement from Herbie's page:  "Error correction in audio CD discs is not perfect; it is algorithm-based "guessing," not binary like in data CDs"
Why is that? and does this apply to, say, McIntosh players spinning discs at double speeds? What about CDs ripped onto HDD?
Any info or links very much appreciated.

For months now I am struggling with my Wadia 781i trying to understand why it refuses to play a few CDs from my collection. Narrowed down, at least some, to pressing defects: some CDs are seriously eccentric, when played on tiny Discmans such CDs make them jump like an unbalanced centrifuge! But only Wadia refuses to play such CDs.
Cros-Interleaved Reed-Solomon Code (CIRC) allows to recover data from up to about 4mm scratches along the track. Between 4-8mm data is interpolated while scratches longer than 8mm cause dropouts - at least for the most of CDPs working in real time (being not able to read sector multiple times). Ripping programs can go over sector multiple times until they obtain proper checksum. I had MAX (OSX) set to "Do not allow to skip" making program going forever with unreadable CD. I had few CDs that I was able to recover that way after couple of hours of ripping. Now I use XLD set to "Max Retry = 200".

I had CDP refusing to play because of dirty lens, then because it needed a little bit of grease to finally stop reading because of worn felt on the magnetic ring that supposed to hold CD down, allowing it to wobble. I had some original CDs with defect - visible wave in internal foil, as well as few CDRs with the same problem. The best CDRs I’ve ever used were Taiyo Yuden. Now I only use HD (server) with a couple of backups.
@kijanki Thank you for responding!! These last few years I got myself a hobby of reviving 1st generation Discmans. Have seen "all of the above" :-(

A few days ago I finally cracked and send an email to Steve Huntley, the designer of my Wadia (he is now with Resolution Audio). His response is below (posted without his permission, of course):
"There is a large variability in CD manufacturing and if the disc is not perfectly circular, or the center hole is off, or the pits are not equally spaced to the center hole its much harder for the player (laser and servo circuitry) to read it. So the lathe (Audio Desk Systeme CD Lathe) is exactly the right solution. Nothing specifically wrong with the Wadia...just picky about the discs it plays. Unfortunately discs themselves are not held to tight standards of manufacture."
"Of course a discman is designed to play anything and while the unit is jumping around all over the place....not exactly the design criteria of a Wadia mechanism. Also units (low cost players) have huge amounts of error correction going on so they can play CD's with peanut butter and jelly on them, kids finger prints, etc! That error correction does not help the sound....the less error correction the better. So again just a totally different design criteria."

What you and SH said confirms my suspicion that when I adjust Focus/Tracking gain, at low settings it skips easily but sounds better. Maybe unrelated, being spoiled by Wadia and Naim CDS3 it is difficult for me to judge the nuances of Discman's sound. 
@kijanki I bought all the books I could find on CD players maintenance and repair, most of them have a few pages on how pits are transformed into sound. If there is anything you can suggest on the theory of different DACs and error-correction, please let me know.

I use Walkman CD players and Walkman cassette players. One thing that Walkman CD player has in common with full size CD players is vulnerability to background scattered light from the CD laser. The background scattered light is a combination of visible red light circa 650-700 nm and invisible near infrared light circa 700-800 nm. The background light is presumed by the rather dumb photodetector to be real signal. The green pen and other counter measures can be employed to amileorate this ubiquitous problem.
I did buy Staedtler 350 Lumocolor marker in Green.
Funny thing, once I treated hybrid SACD of Grieg Piano Concerto with this marker, its SACD layer got busted: starting at track 2 it skips and stalls. CD layer is OK. This proves, I guess, that using green marker does have some effect.
Someone joked that Green is for the Spring/Summertime edge-painting, I shall try flat black tomorrow. 
Whoa! What! Never use green on Grieg! 😀 I’d be interested to hear what happens with flat black. I predict you will not like the sound. But if you use black on the inner edge I predict it will be OK. You can also try purple on the outer edge. A lot depends on what the label color of the disc is. What color is the Grieg SACD?
Problems with players reading CDs is typically the result of the CD itself and not the player. I have over 3500 CDs of varying brands, and many of my disks have been played back on a handful of different players.

I’ve never had a player in the price category of your Wadia though. At the moment, besides my Oppo UDP-205 which is a universal player, the only stand-alone red book CD player I have is an Emotiva ERC-3. One of things it does extremely well is "read" the disk, and in fact the ERC-3 differentiates itself from many other more expensive players because it spins the disk 3 to 4 times faster (1000 RPM) instead of a most players (300 RPM) and what that does is allow for re-reading of the same information.

However, if error correction algorithm is engaged then I also agree this is not "perfect" and is in fact a "best guess". In fact, this notion of "guessing" can be demonstrated when using a PC and the defacto standard ripping software, Exact Audio Copy. If a CD that cannot be read in a particular spot is ripped using "burst" mode, the data (albeit not all of it) can be copied/ripped. If you use a "secure’ mode in an attempt to "perfectly" read the data, often times this will not work (i.e. hang the software) or if it does work take an hour to complete instead of minutes.
First released in 2008, 9 years ago that for a sacd/cd player is getting long in the tooth, it may need a new laser, as back then sacd/cd lasers weren't the most durable.

Cheers George  
However, if error correction algorithm is engaged then I also agree this is not "perfect" and is in fact a "best guess"

Error correction and interpolation are two different things.  System will interpolate when it cannot error correct any more (gaps longer than 4000 bits).  When it is able to error correct then data is recovered bit perfect.  When you error correct you can still get expected/verified CD checksum.
Data CDs are also error corrected.  Extra bytes required for error correction waste about 15% of disk space.  For instance 700MB CD-ROM has 737MB data capacity but in reality contains 847MB.  Additional 110MB is used for error correction coding.

Today I gutted one cheapo Discman, what is left is just a spindle motor and plastic bottom half of the player. Connected spindle motor to variable DC. At 2V / 1000+rpm some CDs make this CD-centrifuge vibrate like crazy (e.g., all CDs from German "Brain" label), some are perfectly centered (4AD label). Wadia refuses to play only eccentric CDs, treatment with AudioDesk CD lathe cures some CDs, and lowers vibration on some.
Oh, and it plays CDR, CDRW and scratch-test CD without any problems

Wadia is practically not used: until a week ago I always used Naim CDS3 for CDs. But I will open it up tomorrow (I did this before, to reset its "gain" switches) to check its CD mechanism.

Cannot wait to get to the fun part of different color markers for CD edges!!!   :-)