There's been a good deal of academic work on architectural differences between open source and closed source applications (basically resulting from the differences in the organizational structures that designed/built/grew them ala Conway's Law). Observations for example include reports that closed source applications tend to have more large scale API classes/layers, because there is a management structure in the designing organization that can herd them into existence, while open source projects of the same size and complexity tend to have a less centralized architecture, again reflecting the organizing characteristics of the developers involved[0].
None of this is arguing that one or the other style of architecture is "better" per se, but rather the architectures are different because they were in the end optimized for different kinds of development organizations.
Most business applications remain fundamentally a three-tiered architecture, with the interesting stuff today tending to happen in how you slice that up into microservices, how you manage the front end views (PHP and static web apps are pretty different evolutionary branches), and critically how you orchestrate the release and synchronization/discovery of all those microservices.
(None of which is directly an answer to your question, but is more meant to say that lots of the most interesting stuff is getting harder to spot in a conventional github repository because much of it is moving much closer to the ops side of devOps)
I'll also mention a somewhat related article here, not directly on topic, but likely interesting to those reading about conway's law and architectures: Microsoft Research did some very interesting work on the interplay of code quality and organizational metrics (e.g. how high in the org chart do you have to go to get everyone who committed code to a specific DLL or what fraction of the developers under that lead engineer committed code to that DLL, or etc). Their conclusion, simply put, was that organizational metrics appeared to better model actual end user experienced shipped code quality than more traditional test metrics[0].
None of this is arguing that one or the other style of architecture is "better" per se, but rather the architectures are different because they were in the end optimized for different kinds of development organizations.
Most business applications remain fundamentally a three-tiered architecture, with the interesting stuff today tending to happen in how you slice that up into microservices, how you manage the front end views (PHP and static web apps are pretty different evolutionary branches), and critically how you orchestrate the release and synchronization/discovery of all those microservices.
(None of which is directly an answer to your question, but is more meant to say that lots of the most interesting stuff is getting harder to spot in a conventional github repository because much of it is moving much closer to the ops side of devOps)
[0] http://www.hbs.edu/faculty/Publication%20Files/08-039_1861e5...