About 15 years ago I devised and performed a detailed analysis of DAE at high speeds. I generated a test signal algorithmically, with values chosen so that any interpolation or holding of data values (so-called "error concealment") could be easily detected. I stored the signal as an hour-long .WAV file, recorded it onto an audio CD, and tested the ability of various drives and programs to extract the data correctly from that CD. I saw cases in which samples or groups of samples were skipped; others in which samples or groups of samples were repeated; and I've seen the left and right channels of a stereo recording get exchanged for small fractions of a second at a time.
If I recall correctly, this was back when the top speed was only 16X or 32X. I saw errors when extracting the audio at only 4X. The limiting factor, it seemed to me at the time, was the ability of a drive to resume reading at the correct point after its data buffer became full. Some CD drives are specifically designed for bit-accurate DAE, and can find that place exactly; the standard back then was Plextor, but they're no longer the only ones with this feature. My tests confirmed that Plextor was doing it right (at least when the company's proprietary DAE software was used), while the other drives that I tried really didn't do such a great job.
(Edited later to add: To understand this problem, please realize that the data on audio CDs is divided into physical sectors--but unlike the sectors on a conventional floppy or hard disc, the sectors on a CD aren't numbered. So "seeking" to a particular sector is only an approximate operation--the player can't tell exactly where it is at any given moment. CD-ROMs and other, later disc formats have sector numbering included as part of the "logical" format of the disc--in other words, the data recorded on the disc includes artificial sector numbers--so that seeks and re-reads can be done precisely. But the audio CD format came first, and it was designed for audio players where the only issue was a few milliseconds' variation in the point at which a given track would begin to be played.)
Of course nowadays, with hard drives that have faster seek times and much larger data buffers, and with CD/DVD/BluRay drives having larger data buffers as well, there might not even be any point at which the CD drive's data buffer fills up completely, as long as other processes on the computer don't interfere. In the optimal case the CD drive can simply keep on reading and its buffer can stay about half-full all the time. Then the drive never has to stop reading and resume where (you hope) it left off. So if I were to repeat this test today, the results might be considerably better, particularly on a computer with a solid-state drive.
Nonetheless, the issue of DAE data errors is very real, and specific measures must be taken to control them. I still use drives that are specifically designed for bit-accurate DAE. Then software such as Exact Audio Copy can still be used for confirmation but there will typically be no re-reads during the actual extraction, because the drive gets it right the first time.
--best regards