I suspect Sony's plan is quite simple, keep cost down and optimise day 1 performance. By sticking with a fixed power supply and thermal dissipation limit, the power and cooling solution can be much cheaper. Less CUs means higher yield, smaller die, cheaper chip. I imagine we will see a surprisingly low market launch price.
With regards performance, suggestions from developers are that it is much easier to gain 100% performance from the PS5 than it is the Xbox Series X. The SSD performance is faster to the point of workable texture streaming, DMA'd into the graphics RAM with a special cache scrubber to reduce the impact. A single RAM speed removes the concerns and need to optimise its use. Good tooling is also rumoured to be a factor. I also don't completely agree with NotebookCheck's take on clock vs performance scalability, it's a new architecture on a chip that has been designed with a high clock in-mind, OK maybe they can't drastically change the paradigm but they're going to have looked at addressing at least some of the bottlenecks.
Most importantly, I think the argument Cerny was trying to make, which I agree with, is release a console on day 1 that is affordable, sleek and quiet, that game developers are going to be able to fully utilise early on. By the time game developers have managed to fully optimise their offerings on the competition and make use of those extra 2TF, as is now the industry trend, it's going to be time for the PS5 Pro, for those who care about performance.