I also agree [[etherpad]] has one of the best user-friendly experiences; we had an etherpad based Agora but in the end I moved to hedgedoc by default as it was more markdown-focused. my ideal would be to run N stoas, all dumping data to file periodically in as lossless a way we can find.