When the new audio analysis with waveform feature was introduced in December 2019 I tried to reanalyze my library. It got about 2/3 complete and then crashed corrupting my library (I lost all views). Fortunately I had a recent backup and I recovered ok, but it was painful because I had to re-enter all of the changes I made over the 7+ days that my computer was doing the analysis.
This experience scared me because as a long time time user with a gazillion hours invested in my library I had never once experienced a crash that caused library corruption. I decided to wait a while for the new analysis feature to mature before I tried it again.
It's now a little over a year later and I decided to try again yesterday. I first cleared all audio analysis fields to ensure all data was created fresh, and to provide a simple method for me to tell which files had completed analysis and which still needed to be done. I proceeded cautiously doing the analysis in chunks of about 20,000 files, stopping after each to make a backup, and restarting MC to ensure all files were flushed to disk.
I succeeded in analyzing about 100,000 files before MC crashed, this time causing a very nasty Windows blue screen.
I successfully restored my backup, and having learned from my experience a year ago, I did not make any changes to my library while doing the analysis, so I did not have to re-do any work.
We can rule out a hardware problem because I upgraded my system in 2020 and have seen the crash on two different systems.
My library may be a little different than the norm because in addition to the usual collection of music and movies, I have a large collection of audiobooks, courses, and podcasts (about 700GB with 3 years of play time). Many of these audio files run several hours per file.
My hunch a year ago, and my hunch today, is that the problem is caused by the waveform feature. I suspect that with libraries like mine, too much waveform data is generated and some threshold is crossed causing MC to crash. The size of my "field (waveform).jmd" file was about 250MB at the time of the crash.
This hunch is reinforced by the following observation. After I restored my library I tried to resume the analysis at my last known good checkpoint by filtering on those files for which waveform was still empty. MC consistently hangs when I try to create this filter, and is apparently not able to handle the large amount of data in the waveform field. I have to kill MC with task manager to exit this hang.
My plan is to continue doing the analysis in chunks, but this time deleting the file "field (waveform).jmd" after each chunk.
Because of the large time periods involved, and risk to my library, I'm not willing to run more tests for JRiver.
I have never used, nor will I ever use, the waveform display feature.
I think you should add an option to turn off waveform analysis, as other users have requested.