Citing Aphyr's analysis of a pre-release version of Redis-Raft as though it applied to a shipped product is so disingenuous as to invalidate this whole post. The very next sentence after the one he quoted is

> We emphasize, again, that these are all internal development builds: Redis-Raft has no production users, so the real-world impact of these issues is negligible.

and of the 21 errors he found, 20 were already fixed before he published his review.