But don't you think it must work
Their own data concedes it doesn't work for everyone and the results vary. That means buying into it is a gamble. That's been a dominant theme discussed by the middle-of-the-roaders here for months.
They may have run this on hundreds of people, but that's still a small sample size, IMHO. Once it's filtered out to thousands or tens of thousands, the real trendline will emerge.
Note the exclusion criteria used to date. I don't know about the newest trial but the previous trial excluded people who had had tinnitus for more than what, 5 years? Duration of tinnitus is a big question mark from my perspective as a long-term sufferer and I'm sure others are also wondering whether it will work for them if they also differ from those in the trials who are now allowed to just buy in and gamble.
Personally, I think the exclusion criteria won't make much of a difference in results, but that's just a gut feeling on my part. Nevertheless, I'd rather be a second-adopter, and I kind of have to anyway, being in the US.
Based on the data we have on hand they haven't established any pattern for why some people respond and others don't other than hyperacusis (something I have, which is a point in my favor). But what I feel should happen (and it's something maybe this forum can assist with) is to gather up volunteer result data and try crunching some numbers ourselves.
I used to belong to the Tesla Model 3 list and they did something similar, they had spreadsheets for reservists and tried to figure out patterns on wait times for cars with various options.
I would assume Neuromod would continue gathering stats themselves so it may be redundant but this would be a way to cross-check their data.