Short answer: yes. Long answer:
Even more CUs could potentially fit, depending what you mean by "realistically". It's a very coarse guide, but say
this block diagram of Navi is roughly accurate to die space allocation of the current 251mm^2 chip. If so, we need to add about 80mm^2 of Zen 2 for an APU. Let's say we also go from two "shader engines" to three. That'd give the chip 60CUs physically. And it would bring the total size to ~435mm^2.
Note that this is somewhat of a worst-case scenario, since it involves duplicating all the support transistors of the whole "shader engine". An alternate approach of adding DCUs to existing shader engines might get you to the same SP count with a (slightly) lower size. But the best-case scenario is probably no less than 450mm^2 for a 64CU part.
Is that "realistic"? Yes, in the sense that chips of this size do get manufactured and put in consumer products. AMD's Vega 64 is even bigger, for example. But it's absolutely not "realistic" in the sense of something that would make it into a mass-market console. It would be too expensive and run too hot.
But on the other hand, posters implying that anything over 40CUs is impossible are too pessimistic. A 48CU part could be as "small" as 360mm^2. This is the size of the original Xbox One APU, and the One X. A slightly larger chip--which approximation from the Scarlett reveal video suggests is the case, at least for Anaconda--could go even higher. (Though some of the physical CUs would almost certainly be disabled for yield reasons.)
I was totally with you until this point. There are various scenarios where PS5 is more powerful, but neither it nor Anaconda are break-the-bank, hotter-than-hell powerplays. Anaconda will be sold as premium, yes, but it'll also be responsible for mitigating losses from Lockhart. PS5 has no such mandate, allowing more budget to go into silicon performance.
That's just one factor making your certainty unjustified. Personally, I still think Anaconda is going to be more powerful than Sony's machine. But that is by no means the only possible outcome (or even necessarily the most plausible).
