"People" aren't really a useful measure here because there is slang that achieves popularity out of the subculture that produces it, subcultures composed of people, just the same as it may not propagate, in which case it naturally remains useful to the in-group. To reject slang as valid language until it achieves mainstream acceptance is a) not measurable as there is no line that is crossed, and b) antisocial.
Anyway, I came up with a better definition: accepting a word/usage is believing that it's a good way to express whatever it was.
I myself will never use "literally" to mean "very," which differentially lowers the chance of this new meaning becoming standardized. Over time, either most people will do the same and the proposed novel usage of "literally" to mean "very" will die out, or else most people will collectively adopt the new meaning.
Empirically, almost all novel usages die out very quickly and are forgotten. Read some slang from the 1920s -- it's already almost incomprehensible to a modern reader.
Indeed, this is by far the usual case. People collectively can, and do, reject almost all proposed novel usages, which are quickly forgotten. We just remember the few cases where things change due to survivorship bias.