That would indeed be fantastic. V8 has had so much optimization whereas CPython has had very little of it, though Pypy deserves a mention here. They've been doing fantastic work, but I guess there's always the curse of the reference implementation being the popular one.
Yes, I'm fairly certain it is. We switched to the SaaS because of that, it wasn't worth the hassle of self-hosting when you get all the responsibility and still pay the same price.
I'm curious if you ever tried MeiliSearch for this, I tried it recently for something unrelated and had a very good experience with it. Since you have the corpus already, it might be worth trying and seeing if it speeds search up? I'd be interested in the results either way.
I tried to use Matrix with a few friends, and while it's impressive how far it's come, it's not there yet. The features are all great, now "all" that's needed is to make them work. We had jankiness like keying problems, messages that wouldn't send or receive, and general problems that made it not "just work".
My personal pet peeve is that I have no idea what the difference is between "people" and "rooms", and why when I try to talk to a friend our chat is sometimes in one and sometimes in the other. Maybe you'd like to take a page out of the Telegram/Whatsapp book and not distinguish between the two, since they're all chats, they just sometimes have two people and sometimes more.
Best of luck, I'm really hoping to switch to Element for all my chats one day!
I did the same, but the only thing that happened was that I couldn't type symbols (quick, type a $ without looking at the keyboard). I exchanged it with one with keycaps afterwards.