That's a secondary source mixing speculation with what AMD said.
"AMD in an official briefing call with us confirmed that the company's "Zen 3" client processors are on-track for launch within 2020."
The way the tapeouts are scheduled means that they could release something in September. If everything goes as planned, which is never guaranteed.
The rumours were that that something would most likely be Milan (EPYC, server CPUs) and that Vermeer (Desktop CPUs) might even be delayed to 2021. AMD denied that, but only said they were on track for 2020. Because the source are the client-segment product managers it is assumed that that includes at least some client CPUs.
We still don't know when and which models exactly. Launches are always spread out over a couple of months so even if you're 100% certain that some Vermeer CPUs will be released in September that doesn't mean we'll see a 4600(X). That could very well take until December.
So stop acting like AMD confirmed a 4600 for September.
The rumours of delays were fueled by Matisse Refresh and yes, that is different from the 2700X Anniversary Edition. The AE was purely cosmetic.
Releasing actually different CPUs, no matter how minor that difference is, in the same segment within 3 months is almost unheard of. Though in this case I can definitely see it being a response to Comet Lake.
Your argument about overclocking makes no sense. What does the 3800XT have to do with this? Why does that mean a 3600XT will overclock worse than a 3600?
You also don't know how binning works. Silicon "quality" is not some one dimensional parameter. It's a combination of frequency, voltage and power consumption. There is some correlation, but a core using less power at the same frequency does not guarantee it can't reach a higher frequency.
And the trick to getting the same power consumption with more cores and near identical clock speeds is the "near".
Power consumption increases massively for Zen2 with every little bit of extra frequency when you're close to or above 4 GHz.
A core at 4.125 GHz needs 14-15W, while at 3.875 GHz it only needs 7-7.5W. A measly 250 MHz extra double the power consumption.
About prices: This is about branding. Navi got different numbers and AMD tried to raise the prices at the same time. It didn't work.
You want the same model numbers (with a different digit for the generation obviously) to stay around the same price. You change how many cores/threads and what frequency the customer gets at that price. That's why Intel "added" HTT to the i5s instead of making i7s cheaper. That's why they added i9s to cover 400-500$ instead of letting i7s cover 300-500$. That's why a 3950X exists, which got 33% more cores and is 50% more expensive than a 3900X instead of selling the 3800X as 3750X. Because x700(X) models in AMD's naming scheme shall cover the low 300s range. That's why the 2700X is called 2700X even though it's faster than the 1800X. Because it doesn't cost 400$.
That is why I can guarantee that there will be a 4900X at around 500$ and there will be a 12 core, but I can't guarantee that they will be the same. Same for 3700(X)/3800(X) and 8 cores.
Better comparision because the 4790K is clocked higher.
So yeah, Haswell needs somewhere between 4.4 and 5.0 GHz for 240 fps, depending on the settings. The lower end of that is definitely doable.
Can you not see the problem? If his current pc could get him 240 avg and a 3600 gets him 270 fps average, but his problem is that he gets drops to 120 how do you expect the 3600 to solve that?
Fakeunlikely to get 240fps from an OC but will hopefully fix the dips he was having.
You said yourself that an OC not even high enough to get 240 fps would somehow fix the dips.
I still don't know how you imagine that would work.