It is a matter of scale. It isn’t just the one server with 10K, like you had on some of the largest IP theft servers. Instead, you have, say, 10 servers with 10K players each in the same building. The problem isn’t with the servers. The physical servers can handle all that and more.
No, the problem is with the connectivity and throughput. In other words, the physical internet connection and bandwidth limitations of the infrastructure, not just in their building, but in the area the building is in.
Let’s use an analogy. We’ll compare the connectivity to plumbing in your house. You don’t like your shower because it doesn’t have enough water pressure. So, where does the problem lie? It might be that there’s a flow restrictor in the shower head which limits water pressure for various reasons. That would be like a physical server which is overloaded. This is something you can fix, either by removing the restrictor, or by getting a different showerhead.
But what if that doesn’t fix it? Well, then you look at the pipes in your house. Some might be leaky, or they might be old pipes, which were run badly and keep the water from flowing properly. This is more difficult to fix without remodeling the whole house, so you may need a new building.
But what if that doesn’t fix it? Well, then you need to consider what happens when you run the shower, the dishwasher, the washing machine, and the garden hose all at once, and whether that affects the available water pressure.
But what if that doesn’t fix it? Well, then you have your connection to the city’s water main. if the pipes coming onto your property aren’t the right size, then you may not be able to get all the water (and water pressure) that you want/need. That’s harder to upgrade, but it is still in your realm of possibility.
But what if that doesn’t fix it? Then you have to look at the neighborhood. Are there too many houses using too much water nearby for you to get pressure? Is the water main old, and hasn’t been upgraded, despite more people moving in over the years?
All of this is an approximation, of course. But remember, the lack of layering/sharding also adds into the amount of throughput required for each character on each server. Say you have 200 people in viewing distance of each other, in Orgrimmar, for a buff drop. With shards of 20 people each, you have the equivalent of 4000 people’s actions streaming to and from servers (each person sees the actions of themselves plus 19 others). Without shards, those 200 people now account for 40000 actions constantly streaming. And that’s just their movements. That doesn’t count the environment, people AoEing, etc. Just 200 people. Now expand it to a server of 10K. And put it in the same building with 10 other servers of 10k.
The problem is scale. Things that work at small scale (and yes, the IP theft servers were small scale compared to Classic) do not generally work the same at larger scales.