adder, welcome back. Good to see you posting again. It's late at night for me, and this is potentially a more complicated discussion than there's time for right now to do it justice, but here are some initial thoughts to get it going. Thanks for such a well-formulated post.
First, I think that in a way you've answered your own question. Your last point touches on what I feel is probably the only answer that would be including the algo changes subsequent to improving the site in the months after Feb 7th, 2017... In particularly, the March update (aka "Fred" was said to be harsh with sites with not very good backlinks.) .
His CF/TF is waaaay out of balance. Ok, ok, I know this is almost as bad as talking about PR, however, I ran a quick analysis of 100 sites within that niche and statistically, the ones with off-balance CF/TF tended to rank badly, his site among the bad pile, of course.
I also don't think that CF/TF is such a bad thing to look at, considering what else is available at this point. Even though TF is from Majestic...not an actual Google metric... and there's lots else in the Google ranking algo, I think I'd trust TF (Trust Flow) more than I'd trust the old public reports of PageRank, to give me a quick reading on backlinks. Majestic TF does break links down into which are meaningful and which aren't.
Also, in the heavy days of Panda and Penguin that recovery from a link "penalty" wasn't automatically going to get you to the top again... It was simply going to get you back to the base level of where your "good" backlinks were without your bad backlinks... and if those pre-penalty links of a site weren't good, "recovery" wasn't necessary going to restore your old rankings, which had been gotten by bogus link votes. Whether it was shaking off either penalty recovery or simply bad backlinks, a site would need to earn new backlinks, and if backlinks got built with the same techniques that got the site penalized originally , then it probably wasn't going to recover.
This may be where the site is now... It may be that the site hasn't since been aggressively marketed enough, or that the site simply isn't as engaging as it needs to be... or that there hasn't been enough time to rank since the site has been rebuilt.
Regarding the early drops....
from what I can gather, this was likely to be a content/UX quality update, although there hasn't been any confirmation.
There's been some contention as to what was going on. Several of the studies reported were quite specific that some Panda-like observations were observed. Google, IMO, was hitting sites with multiple algos, I think, perhaps to obfuscate something else that was going on... or perhaps they were running extended tests on a subset of sites that, say, were outliers, or sites in spam-prone niches, where Google needed to rely on combinations of signals that didn't hit all niches. I don't know. I haven't seen this kind of research on the algo, and Google hasn't said much.
The idea of being cautious to avoid excessive monetization efforts, one of the vulnerabilites mentioned in various studies, is a factor that I can see being tested with an additional algo "level", apart from Panda, and conceivably it might have resulted in the kinds of algo cycling which were reported at the time.
It may also be that linking in some areas changed to need more topicality to be effective. This might account for the very precipitous drops, which remind me a bit of "Florida". I should note the John Mueller, after the February change and then "Fred" in March, suggested that much more than small quality fixes would be required in many sites he saw. In some niches, he seemed to be saying that it could well take something astonishing and new to get searcher, linker, and Google's attention.
Anyway, I toss these ideas out... in some cases educated guesses, and in some complete conjecture. I'm relying mostly on memory about the algos.
More from you about the background of the site and how competitive the niche is could be helpful. I hope you keep us posted about how this evolves.