What I mean is that OP's question and your answer both have a context which isn't always explicit in these conversations.
The Linux kernel development, and I would argue most successful remote shops, have had a "remote first" culture from the beginning, or, if remote was something that came up later, have had a change in culture.
My client, for example, has a culture where people would just drop in and ask random questions. Or you would overhear people talking about something and you'd jump in. During lockdown, this wasn't possible anymore. So they would start having meetings among "stakeholders", which, in the above story would be the people initially in the conversation, and reach some sort of decision. The other people wouldn't be able to jump in during the conversation. They could of course argue with the decision, but sometimes they'd find out about it "too late".
Contrast this with a culture where everything is done in public, for everyone to see and with a possibility for everyone to intervene during the discussion, as is the case on the LKML or as I suppose is the case in a remote-first company. There would be discussions say on slack or some other communication tool where even people not directly invited could at least have a look and see what's emerging.
I think that those people "wringing their hands about how there’s no way to do spontaneous, creative, collaborative work remotely" are people used to working in in-person cultures. Not only that, but it's very likely that the companies where they work are very "in-person" companies and that those companies have a hard time adapting to the remote way of doing things. It's not enough that some people would try to make this work. For this to actually work, there has to be a serious change in how the company itself works.