I may have a solution for those issues in the coming weeks.
Lower disk space (about 1/5th), significantly lower RSS (9x, close to 10x), significantly lower write strain on disk (over 500 times in fact during writes, haven't benchmarked read yet), lower CPU usage (currently 10-25x, might change), faster queries (8x), significantly faster ingest (200x).
To be clear the software exists already so I'm not speculating, but it needs a lot of work and improvements.
