Do you have any statistics to back up that claim?
I realize that hard statistics are difficult to come by on this matter since Apple doesn't publicly disclose the failure rate of their keyboards, but one of the closest things we have to "official" statistics is
these statistical data from a former Genius Bar member that circulated around the Apple news sites a while ago, and it suggests that the failure rate of the keyboards in the 2016/17 MBPs is roughly twice as high as with the pre-2016 models. Which, without a doubt, is worryingly high, and at least twice as high as it should be (since issues like that, from a consumer-point of view, should never become
more frequent than with a previous iteration, rather
less so).
This iFixit article about the Butterfly v3 keyboard, while admitting that it's still possible for dust debris and especially sand to sabotage a key (so it's not a perfect solution!), suggests that significantly less dirt can get below the keycaps and that these membranes do an overall fairly good job at what they're supposed to do. It doesn't give us any concrete percentages, we can still put one and one together: if the rate of debris-caused failures on the new keyboard is even just somewhere around 50% lower than on the butterfly v1/v2 keyboard (which, by this iFixit article, sounds highly likely), then, by the previous statistic, it should be around 2015 MacBook Pro-levels. It's very possible that it's still slightly higher than that, but at the very least, the failure rate of the 2018 MacBook Pro keyboards should be much closer to the failure rate of the 2015 ones than the one of the 2016/17 ones, based on these two statistics that are likely as good as statistics on this topic can get.
If you have better statistics, then by all means, I'm happy to be proven wrong. But unless that's the case, it seems to me that your comment about my statement about the keyboard failure rates in the 2018 models to have significantly declined "simply not [being] true" is... well, simply not true. I'm not denying that there are still issues, but if the keyboard failures on the ≤2015 models seemed to be half as frequent as on the 2016/17 ones like these statistics suggest and nobody considers them an issue there, then I don't see a good reason for why we should treat the 2018 models all that differently from them when their failure rate should be somewhere in the some league.
I think this is just bad advice.
It's bad advise to advise people to buy an in many categories far superior product if the price difference is small enough just because of a slightly higher (but, again, statistically far less than twice as high) failure rates of the keyboard? A product that very likely also eliminated or at least improved other points of failure such as the Staingate issues?
I mean, to those people who are
that afraid of a slightly higher keyboard failure sometime down the road, yeah, I might advise 2015 models. Or if you absolutely hate stuff like the Touch Bar, or want way more power than their thermal design can provide. But I'm not sure if throwing under the boss the numerous and in parts fairly significant improvements of the newer models in the (not necessarily successful) attempt to be slightly more foresighted in one particular area is that clever since you're giving up so much for it. If we're talking about a 1000€ price tag for a 2015 MBP and a 3000€ one for the spec-wise comparable 2018 one, then sure, go with the 2015 one instead. But if we're talking about, say, 3000€ vs 3400€, which was more akin to the MSRP price difference back when Apple still sold the 2015 models (which was only until a few months ago), then my general advise to the majority of people is to rather byte the bullet now and go with the machine on which photos and videos (and everything else you do) will look far better, that is leaps and bounds ahead in sound and speakers, that is way faster in pretty much every area, is way more portable, has Touch ID to minimize your password-typing, a far larger trackpad for more finger- and gesture-freedom and so on and on.
Assuming Apple adds the 2018 to the 4 year keyboard plan, and assuming you aren't keeping it more than 4 years, sure, go ahead and get the 2018 if you are cool potentially dealing with losing your machine at times for repairs just to replace it with the same defective keyboard again. But if you are keeping it more than 4 years, the 2015 will be a better buy than the 2018 from a reliability standpoint. I would expect 5 more years of MacOS support, if not more, because my mid-2010 is supported through High Sierra, which up until a month ago was the current OS. There is plenty of life left in a 2015 machine to last well to the point that Apple figures out the keyboard with presumably a 2020 redesign.
I actually did just that 4 months ago. Since everyone else here seems to share anecdotal evidence, let me tell you my experience. In terms of mobile Macs, I currently own a 2016 MacBook (12") and a 2018 MacBook Pro, and I owned a 2014 MacBook Pro until a couple of months ago as I sold it to friend.
Want to take a guess which one of these three I had to bring into an Apple-autorized service provider for ~2 weeks to get it fixed as part of a replacement program? The 2014 one – namely, because it suffered from some pretty severe display stain issues, so bad that even when using the display at full brightness and with bright colors, the outlines of the stains were visible on-screen. The first time I went to that reseller, they even told me that they wouldn't fix it for free because it was out-of-warranty and the old staingate extended-warranty program was just over, and offered me to replace the display for around 500€ instead. I declined – fortunately so, as the program was continued just a few weeks after that, and I got my free replacement after all, but for some weeks, I was faced with either having to live with a grotesque, distorted, always-visible Bat-signal on my screen, or with having to put down a couple hundred bucks (and weeks without my machine) to get it fixed, even though it was even still less than three years old (no AppleCare). I don't see how my experience there is in any way better than what people who had their keyboard fail on them had to go through.
Moving on: Want to take a guess which one of these three machines suffered from kernel panics 1-2 times per month on average, sometimes more frequently, without me ever finding the source of the issue? Once again – the 2015 model. Now it wasn't frequent enough for me to take it in to a repair (I mostly just noticed it after the warranty was expired anyway, it didn't really happen at the beginning), but there were still a few times where it caught me at a rather inconvenient time. I'm not denying that the T2 chip can be a cause of kernel issues, but they are most certainly not the only possible cause of them, and you are not immune from them on older Macs either.
And these were only the two most severe issues – there was more, for example sometimes the display would flicker or even freeze briefly, or even just show a frozen screen for a few seconds, after waking the Mac from sleep. I later read on the web that the most likely cause for this was a "loose" display cable somewhere inside, which is reinforced by the fact that this particular issue was gone after the display replacement due to the staingate issues. The display also developed some fairly noticeable backlight bleed over time – something that, again, was fortunately "fixed" due to the staingate display replacement, but that I would otherwise have had to live with. There's other minor stuff like that the LED on my MagSafe 2 cable eventually stopped working, or that the MBP sometimes wouldn't start charging after plugging the MagSafe cable in and I had to dis- and reconnect it, and probably more that I don't all remember. Ironically (or fittingly, for this thread), I've also had some slight keyboard issues, though not with some keys malfunctioning, but with one or two keys sounding/feeling a bit weird when hitting them (possibly due to debris under it?), and one key also lost some of its coating (but that might be my fault from... I dunno, typing too much).
To circle back to my original point: do you want to take a guess how many issues I've had with my 2016 MacBook or my 2018 MacBook Pro? Well... not really anything that I can think of. I mean, yeah, they are not quite as old as their oldest brother, but I'd be lying if I said that I had any serious hardware problems with either of them. The 2016 MacBook has maybe some slight backlight bleed but nowhere near as serious as the 2014 one. I also have a 2014 iMac that has had some issues, but overall, the 2014 MBP was by far the Mac that I had the most trouble with to date, whereas I really don't have anything in terms of issues to complain about my two newer MacBooks, and my 2016 MB turns 2.5 years old soon.
I'm not claiming that my experience is universally shared – judging by the responses in this thread, it's certainly not. But the 2012-2015 MBPs are not without their own quirks and points of failures, some of which very likely have been improved since. Are the 2015 MacBook Pros still the most issue-free iteration in MBP-history? Well... probably, all in all. Many people seem to think so. But issue-free, they were not, and just because some release has slightly more issues than it doesn't mean that you should avoid it completely. That's my opinion, anyway.
I also highly contest that the more than 3 years older model is a better choice from a longevity point of view – the oldest currently supported MBPs in Mojave are the 2012 ones, so it's not unreasonable to guess that the earliest the 2014 MBPs could become unsupported software-wise is 2020. Now in practice, I don't think it will be that soon – I'll rather expect another 3-5 years, unless Apple does some serious cleanup in one of their first ARM-releases of macOS. But you'll still have to consider that the 2018 MacBook Pros might easily be supported twice as long or even longer, especially when you consider that in the iOS world, significant performance leaps (such as with the 2018 MBPs) or the introduction of some completely new SoCs/architectures (such as the T2 chip in the 2018 Mac lineup) oftentimes became the cutoff-point for the supported OS versions. My prediction is that especially the 2018 MacBook Pro with its fairly significant internal changes will be a far better choice in terms of "future-proofing" than the 2015 whose keyboard might be slightly less likely to fail. But I guess we'll see about that in the next couple of years.
I'm, by the way, also doubtful if the 2020 redesign will really bring a completely redesigned keyboard to the table. My guess is that Apple will rather improve the Butterfly design and the membrane mechanism (and make it more failure-prone than it currently is). Knowing Apple, they won't back paddle to a thicker MacBook Pro (at least not so much thicker that they could fit the old keyboard inside – no need to tell me that the 6S was technically 0.2mm thicker than the iPhone 6), and if they wanted to abandon the Butterfly design, you would think that the 2018 MBPs (or the just-released MacBook Airs that feature the same keyboard!) would have been good opportunities to start over. Remember that the Butterfly keyboard is already almost four (!) years old at this point (originally debuted in the MBs in April 2015), so by the very-same "redesign each four years" argument that people use to predict a 2020 MBP redesign, it's not unreasonable to think that Apple could have put out a redesigned keyboard this year if the Butterfly one was really un-reedemable.
Well, judging by the thread on this very subforum, butterfly V3 didn't fix anything.
I don't think we can properly determine the frequency of an issue by a thread on which almost only people who do have the issue comment. I'm not denying that there are still issues, but by the very statistics I gave at the beginning of this comment, aswell as by the consensus among most reviewers of the 2018 MBP and MBA, I have a hard time believing that the Butterfly v3 "didn't fix anything" and that the issues are still anywhere as bad or widespread as with the previous two iterations.
