1989 rules, CL p 11
1 hit/ hour resting
1 hit/ 3 hours active
which is the origin of the odd /3 hours /hour /3 hours time frames likely.
Sleep isn't mentioned in CL
There was originally no effect of Co on healing rate, so the backward healing of negative Co bonus is not an issue.
As to why the RMFRP crew chose to change the factor to Co/2 or Co*2 and not also normalize the hourly rates I dunno.
Peter's Fix is likely the right direction:
I've not had a problem with the RAW (except for Co bonus 0 or lower, which we treated as +1):
Active = 1 hit / 3 hours
Resting = 1.5 hits x Co bonus / 3 hours
Sleeping = 2 hits x Co bonus / 3 hours
The only reservation or caveat I'd put into that, is that the 1/3hours 1/hour was intentional. . . sitting down to rest for an hour does you good, you don't need to rest for 3 hours to get a benefit. . .but if your sleep is interrupted and you don't get at least 3 hours of it, it's really just resting, not sleeping. So it's possible that what looks like a mistake was done intentionally. . .and while peter's 2nd variation of 1/1.5/6 maintains the 1:4 ratio of resting:sleeping, it doesn't catch what's essentially a 1:3 ratio from active to resting. . .and 1/3/12 per 3 hours active/resting/sleeping might be a bit extreme as a solution.
Keep in mind that the stat set ups are fairly inflated for PCs, and the races have +0 to +6 Co bonuses, so likely they rarely had negative Co PCs in play, and out of play, out of mind. . .it's the sort of problem that comes up more often in the hypothetical than in play. (Similar to the x2 St from the Strength spell. . .which makes a character with negative ST bonus weaker).