in general, it looks to me like each observation generally needs ~1.25 identifications from others to keep status quo. you could think of that as 0.25 to get an identification on observations that are never identified by the observer (1 out of 4 observations), plus another 1 identification by others on average on each observation. status quo would mean that 62.5% of verifiable observations then reach research grade based on that.
of course, if you want to do better than status quo, you need more identifications – either more from others, or more from the observers.