Mixing 4000-series CMOS and 74HC in a 5V system - any issues?

That's where I find it, absolute maximum ratings. You are right in that it has no place in recommended operating data.

Correct, this is a little something just for fun. All my serious designs use rather exotic monolithic processes with full dielectric isolation, complementary bipolar with SiGe, and CMOS. That's a different kind of fun, but it pays the bills.

If the entire circuit is CMOS I don't see mixing the two families giving you any trouble. The issue is noise causing double clocking on a slow edge. Are you generating any clocks from logic? Are you driving any clocks from a slow edge? If not, don't sweat it.

With the experience you have I guess you just aren't used to low power, slow logic, eh?

I've done my share of low power and logic stuff too, but it's been either monolithic or all in one family, with edge rates pretty much the same across the board, subject of course to loading and proper drive sizing. But mixing 74HC and 4000-series is a different kettle of fish. HC edge rates at 5V are two orders of magnitude faster. 4000 logic was introduced decades before 74HC, so fast edges were never really a consideration. My main concern was whether there might be something in these designs that had problems with significantly faster edges, info that might not have been known at the time and thus wouldn't have made it into the old RCA datasheets that continue to be copied over the years. Given the basic logic design I wouldn't expect that as a problem, but I thought I'd ask since it could be painful to debug.

I'm guessing they're a little vague about what the absolute minimum voltage where the part won't be damaged but stops operating properly is because the manufacturer doesn't know themselves for any particular instance of the part; if you could make an "ideal" MOSFET with an infinitesimally small threshold voltage and on resistance then a CMOS gate should work fine with the rails arbitrarily close to each other.

So there's some distribution of real-world on resistances and threshold voltages for the process they're using to make the chips, and 2 or 3 volts minimum "recommended" respectively is just the threshold where they can statistically guarantee from population extrapolation from a random sample that 99.9....% of instances will operate as expected.

Edge rates by themselves are not a problem. The problem is created by noise in the circuit. So it's not so much a problem of mixing edge rates as it is using a device with a slow edge rate and a device that generates noise. Control your noise and the edge rates won't matter. Even if the logic was all 4000 series, the slow edge rate will be more sensitive to noise than a faster edge rate. Bottom line is to control the noise to a level that won't mess up operation of the clocks. The data signals won't matter.

It may be an issue of "don't know" but more likely an issue of "don't care". Anything they put in the data sheet has to be verified somehow. If customers don't care about that feature why bother to measure and publish it?

Bingo. The real cutoff may be much lower, but the numbers they provide cause them no heartburn and makes most of their customers happy.

I recall seeing memory devices which were specified down to exactly 1.8 volts. But to use it with a 1.8 volt supply you would need a lower voltage spec for tolerance on the Vcc. I figured they must not have much demand for 1.8 volt operation but wanted to add it to the data sheet anyway.

It doesn't mean that. If you reverse the rails on a CMOS chip by more than 0.5V you can put enough current through the chip to warm up some areas to destructive temperatures.

A more explicit warning would have read "don't put significant reverse voltages across CMOS supply pins".

I would include current limiting series resistors on all 4000 series input pins.

The 4000 series had a tendency to latch up and let the magic smoke out if the ESD diodes conducted a bit too much, even on a transient.

Maybe that was the 4000A series and they fixed that in the 4000B? I would still add the resistors.