I would then make an assessment of the veracity and usefulness of the test. I've been building PCs longer than Puget Systems has been around (not by much, but a few years) and I've been using Lr since it was released. I'm not comfortable with their assertion because it doesn't seem to marry with real-world results that I've seen, but I'm very much prepared to accept their testing if it's robust and measures real-world scenarios regardless of how uncomfortable the results may be to me because good testing is always better than general observation. I'm always concerned about claims and tests that don't detail the methodology, though.
And I was involved in product development and product management of the PC for 20 years starting in 1983...and started with Lightroom on the first beta....but that does not make me a testing guru, nor have I performed any concise tests of my own.
However, their conclusion statement seems reasonable based on my experience in using SSDs vs HDs with LR and statements Adobe has released on performance considerations.
That statement is “...Whether the source images themselves were on an SSD or a platter drive did not appear to make all that much of a difference, but having the catalog, preview, and cache files on an SSD allowed us to convert RAW images to DNG about 2% faster and export images about 7-8% faster. Upgrading to an even faster NVMe did not further improve performance very much, however.”