I think you have to look at the products themselves with regard engineering. There is very little to go wrong with a direct drive motor hub and they are typically very overbuilt and heavy however some are coming out of fairly basic Chinese factories but still overall a very reliable and simple motor. The geared hub have the addition of the 3 planetary gears and clutch system but still very simple but it does add in a wear and tear element to the design and then you get to mid-drive which is a very complex solution in comparison and has to deal with not only the motor power but the riders own power going through the gearing of it. It is extremely complex compared to a hub motor. Some mid-drive motors have belts and the controlling circuitry is within the motor assembly and water seals do fail on occasion. However typically mid-drive motors are manufactured by companies with higher engineering standards like Shimano, Yamaha etc. It would be amazing quite frankly if mid-drive motors could achieve the same reliability as hub motors. People buying hub motor ebikes may not be bike enthusiasts they just buy the bikes probably mainly for commuting and ride them and that's it really.I suspect there aren't that many people who have proper experience of both mid-drive and hub drive, and purchase decisions are made on other criteria - price being a major one, A mainly cost-driven decision will tend to result in a hub-drive purchase.
I have never ridden a hub-drive - although I will when my conversion kit for the Brompton arrives. Meanwhile I continue to be astonished at the natural feel of the Bosch system, especially in Eco mode. 40% supplement to my own effort, if that's what it is, is transformational for me and so far has got me up all the nearby hills without me being prostrated through shortage of breath, using the appropriate gear of course. And it actually feels as if I am doing it myself, which I know is not the case.
I don't expect to be disappointed with the hub drive, but my expectations are that it will be a quite different experience, particularly without a torque sensor.
As to reliability - I doubt is there is much reliable and sufficiently detailed data in the public domain. I have some experience of fault and breakdown analysis in relation to domestic appliances. Shame it wasn't bikes
What you need to know is both the frequency and severity of breakdowns, and to have a fault cause analysis. Some brands have lots of mainly minor faults, some have very few but on average are very expensive to fix when they break down. The most interesting analysis I recall related to a white goods category. In both total repair costs, and frequency of breakdowns, the 'worst' brand was four times as bad as the best ones, and twice as bad as the average; and 80% or 90% of the breakdowns on that make had a single cause (identified by fault code) which, had it been eliminated, would have made that brand one of the most reliable.
The products in question were in the market 30 years ago so it wouldn't be very helpful or fair to reveal which manufacturer made them, but the point I am coming to is that until the analysis was done nobody would have guessed that the actual reliability of a well known, high volume brand was so (relatively) poor. The make was generally perceived to be OK - and had a high market share. It helped that the repair network was very good (and they obviously got a lot of practice). Good product support has a big effect on reputation, and sounds like an opportunity for Bosch.