Why are bigger blocks dangerous for bitcoin?

If you are like me you probably are in a state of confusion about the scaling debate, especially with one side of the argument seemingly not even interested in making sense.

First things first, I'm not a computer scientist, cryptographer or mathematician, so be critical of what I say and use enough salt and do your own research.

Pretty much everybody even remotely familiar with cryptocurrencies should at least be aware of the ongoing debate around blockchain-scaling with bitcoin taking center stage. (Btw, am I the only one who thinks it's weird how eth-fanatics are seemingly oblivious about the exact same problems on their own platform while they are making great effort pointing out flaws in others?). On the one side there is this imo dubious and incoherent effort by bitmain and Roger Ver, that are the main driving force behind the unlimited blocksize argument. They seem to be convinced that bitcoin-core is centralized in the current form. So they are now going to centralize btc with a president (did somebody not realize that position was just supposed to be a running gag), new developers in their control, censored dev-repository and other nonsense all in order to "save" bitcoin, all the while preventing peer review and apparently planning to release that untested and unvetted code directly onto the main-bitcoin-network? Hard to believe that anybody thinks this is a good idea.

While I've seen a lot of people claiming that bigger or unlimited blocks are just fine, quite a few experienced btc-developers seem to think it is dangerous. So far I have not seen any evidence confirming the claims that big blocks do not negatively affect decentralization. Actually even people put forward by camp bitmain not only confirm the centralization aspect, but also seem have explicitly set this as the goal. So skipping over the obvious risk of Jihan Wu, wanting to change bitcoin to his own ludicrously inefficient personal paypal-like monstrosity in a communist regime controlled jurisdiction, I like to point out two other risks of big blocks I ran into, that aren't discussed a lot as far as I could see.

I lost the link to this post and have been unable to find the source again, I still list it because it seems to be a valid concern. (If someone has a link to this article, posting it in the comments would be greatly appreciated) I remember reading someone doing the math on bigger blocks in combination with a cluster of miners with a delayed connection to the rest of the network. According to this math a cluster of 30% hashpower with better interconnection to each other compared to the rest of the network would end up taking over entirely. I suspect this is what Luke-jr is referencing when he's wrote that even 1Mb blocks are potentially dangerous.

Segwit2x sudden introduction of 8Mb blocks will result in 90% of the nodes disappearing according to this study of bitfury. Meaning a sudden increase would be bad. I'm not saying that the blocks should not increase, and just about all bitcoin-core devs are of the same opinion afaik, but they want it done in as reliable and tested way possible.

The way I understand it, bigger blocks alone do not scale linearly, will not stop spam, do not actually solve scaling and will need off-chain-scaling-solutions anyway. Also segwit not only seems to increase blocksizes, but makes transactions smaller and more efficient to process.

Anyway I'd love it if people could prove me wrong or give me extra information on this in the comments. Like I mentioned I've been having trouble figuring out what the dangers were that bitcoin-devs keep referencing, aside from the obvious insanity of running untested code on the main btc-network.

To me seeing the miners blocking segwit and attacking the btc-developers, but then embracing everything they made anyway seems to be about something else entirely. It looks more like a powergrab then honest concern about btc. The following meme on bitcoin-reddit seems to hit pretty close the mark from what I can see (click the image for the reddit-thread):

Meanwhile have a nice weekend all!

PS
Just as a heads up, I've seen a btc-developer write that juli 21st would be something of a deadline to be on time to signal for segwit, after that forking might actually become a thing.