Classical titles are a whole different - and much more challenging - thing when it comes to metadata, for sure.
As far as reliability, years ago when I first started ripping CDs I used iTunes. And I found that a small but worrying percentage of the CDs I'd ripped would exhibit audible glitches. Almost always these were short bits of just one or two tracks on a CD.
This was all due to the lack of sufficient error correction, and the way iTunes has a drive read a disc.
I soon switched to secure rippers - first a now-discontinued app called Max, and then to XLD. Both were/are freeware, and both are for the Mac (which is what I've always used). These apps have the drive read a small segment of the disc multiple times until they get the identical result at least twice in a row, and then move on to the next little segment. They also query the online AccurateRip database, which contains rip logs and hashes from millions of other people's rips - this adds an extra layer of confirmation as to whether the ripped data is correct or not.
Comparing rips using these apps with rips of the identical CDs I'd previously made with iTunes, I found that my iTunes rips fell into four categories:
- Audibly bad iTunes rips that were due to unrecoverable flaws on the CD;
- Audibly bad iTunes rips that were due to easily recoverable flaws on the CD, fixed by a secure ripper doing its "read multiple times until we get two identical results" thing;
- Audibly good iTunes rips of CDs that the secure ripper was also able to rip error-free;
- Audibly good iTunes rips of CDs that the secure ripper found unfixable errors in, but the secure-ripper was able to complete the rip and the errors were few and tiny and impossible to hear during playback.
With the secure ripping apps I have found that about 96-98% of my 500-disc collection can be ripped 100% error-free (this does include discs where the app reports having to do some retries on some tracks, but that's not unusual and the end result was a perfect rip). Another 2% or so will rip with errors, but when I grab a confirmed 100% perfect rip of the same pressing online and run a null test in Audacity, the errors in this 2% of my collection consist of tiny glitches involving a handful of samples in a couple of isolated spots, and are totally inaudible.
The final 1-2% of my collection is - or was, since I've replaced that handful of CDs over time - discs that could not be ripped 100% error-free, and whose glitches produced audible issues or prevented XLD and the drive from completing the rip at all.
These years of secure-ripping experience and examination of waveforms in Audacity have also turned up all kinds of weird little discoveries. The one that always surprises me is how often when you compare rips of two different CD pressings that were made from the identical digital master, the rips don't match up 100%. The culprit is digital sync errors - one of the two CDs will have one sample repeated every 1-2 seconds or so. This is completely inaudible, but what makes it especially bizarre is that over the course of an entire album, the one with the repeated examples will have a very slightly longer run-time as a result! This is just fractions of a second over an entire album of course, but it's bizarre.
Finally, there is one other major benefit of the secure-ripping apps over something like iTunes: drive offset. XLD and I believe other secure-ripping apps allow you to set the drive offset - how many samples the drive automatically reads ahead of, or after, the official start or end of a music CD track as per the CD's table of contents.
I don't completely understand how this works or why it's even a thing. But with software that makes its own, unchangeable, behind-the-scenes assumptions about drive offset, you can end up with a situation where a tiny but audible bit of the beginning or end of a track is cut off when you rip it. When you rip a whole album, the effect is that the first or last fraction of a second (or in some cases more) of each track is actually part of the prior or next track. It's not an issue if you play the entire album through using gapless playback. But if you play just one track it becomes a huge issue. In the early, wild-west days of the music file-sharing craze in the early to mid-2000s, tracks with this issue were all over the place.