I think the UI mostly makes sense, it has a hierarchy from left to right, so it is essentially a amped-up tree-view. You could argue for actually using a tree view (a la TeamSpeak), at least on desktop, and I wouldn't disagree, but it kind of makes sense given parity with the mobile app.
I don't see a need for minimal input delay for text, compared to a code editor, as when writing natural languages, they are mostly formed in sentences and therefore have a lot of "buffer" in my brain, I don't need a closed feedback loop to type. Suppose that someone that relies on that would be rightfully befuddled with any delay.
It uses 156 megabytes of memory even on this high-volume scenario, so I would say that while baseline performance is low, it doesn't get slower, which is good.
That being said, I can always find what I want in the UI, it is rather internally consistent. The voice features are excellent and it has less outages than Slack, in my experience!
