> The ground and the sky exist.
Those are directly observable. "Fact" stands in for an "observation", or maybe a repeated/repeatable, perhaps standardized, observation. (The sky is really just the way the atmosphere appears from a certain vantage point, so if you really drill down I'm not sure what saying sky itself exists actually entails, but I digress).
> Everyone can perform an experiment to verify it.
Sure! But in the modern scientific paradigm the purpose of that experiment would be an attempt at falsification. This is absolutely critical to the underpinnings of science — anything that is inferred from observations could be falsified at any time, so it doesn't really make sense to allow a secondary inference — a hypothesis or theory — to be linguistically elevated to fact, which most people would take to mean something that is concretely true. I can understand using fact to refer to a primary observation, which is where fact seems a natural word. I'd still argue that observation is a more apt word, especially as they often involve instrumentation or cognitive bias and are rather subject to change.
I'm not expecting atomic theory to be falsified anytime soon, don't get me wrong. But in only slightly less edge cases, like round vs flat earth or evolution vs creationism, terming the beyond-a-reasonable-doubt consensus a fact isn't doing any work. Either we agree, at which point it isn't really necessary anyway, or we don't agree, because someone in that collective we is a flat earther or creationist or similar, at which point just using the word fact isn't going to convince them anyway.
> It makes communicating scientific insight harder.
On less epistemological grounds and more pragmatically... I do a bit of sci comm of my primary research, some public facing research and outreach and have also spent plenty of time arguing scientific consensus topics on the internet with people who may or may not see them as such. Anecdotally, most of the time I see the word fact being bandied in the context we are discussing here it is used as a cheap stand-in for where some actual evidence or at least evidence of scientific consensus should have been. (Quite often it is invoked in an attempt to shore up something on shakier foundation or for which the invoker just isn't that well versed in and may not be correct about.) Just not doing much work.
Take the atomic theory example — to teach it the first time, depending on the age group you would just describe atoms, without ever needing to stop to call them factual, or you would provide historical experimental and other evidence of the existence of atoms and why they are a useful model for physics and chemistry. The word fact isn't necessary. If atom truthers emerge and you are trying to convince an undecided or torn public that they really do exist despite what the YouTube videos say, the word fact isn't going to help.