The many eyes theory is broken due to complexity and amount of code, and hence the primary way to the future is in reduction of both (loc and complexity thereof). At over 10 million loc these days, linux has become unweildy for even the most experienced kernel hacker to really understand. This is why I think microkernels like Minix 3 have a bright future as their development progresses. (Minix 3 has <15k loc)
Here's the problem: At what rate do microkernels like Minix 3 progress? Compared to Linux, they progress glacially.
And, when Minix 3 has as much functionality as Linux does, how many loc will it have? I grant you that you might have some efficiency from the (assumedly) better architecture, but how much? Would Minix 3 have 5M loc? Would it be really understandable at that point? (Perhaps more than Linux, because of more modularity. Still, you're essentially comparing a robust product to a nowhere-near-production-ready proof-of-concept. It's very much not apples-to-apples.)
I completely agree, but one thing to keep in mind is that Minix 3 just relatively recently got it's BSD license, and I think that was it's main limiting factor in the past. As for how it progresses, I can't predict the future, but I think something similar to it could easily get a good boost if it was embraced/pushed in the sector it's best at, eg embedded devices and iot. I work along-side scada systems and am constantly appalled at the state of security on systems.
The many eyes theory is broken due to complexity and amount of code, and hence the primary way to the future is in reduction of both (loc and complexity thereof). At over 10 million loc these days, linux has become unweildy for even the most experienced kernel hacker to really understand. This is why I think microkernels like Minix 3 have a bright future as their development progresses. (Minix 3 has <15k loc)