more OT: Just now I am using it for develop some simple DCV 10V transfer standard so that I can "pic-up" accurate 10V from "lab".
(reference chip and around components are all in tightly temperature controlled small oven (just like good OCXO but low temp)) I hope (try) get trusted +-1ppm or better for short time (I give "hands up" for +-0.1ppm). Noise is one problem but noise is also random... so not very big problem. But random "jumps" are more problematic.
The LM 199/299/399 (discontinued) was used in lots of bench meters, it's a stable zener (voltage is something like 6.9V) with an internal heater that keeps it at a constant (high) temperature (like an OCXO), so the tempco is very good. It's still available on ebay. The LTZ1000 is newer, very expensive, and needs more supporting circuitry. I believe the noise perfomance is pretty variable, some devices are good, some not so good. This doesn't take care of the tempco of any dividers or amplifiers. I would age them for a few weeks or so to skip the initial drift, according to the HP Journal article, they aged them for two months (and probably selected them for drift). There's also someone selling HP 3458A reference assemblies for something like $85, he does age them for a number of hours. Not sure how easy it would be to use these.
Rigol DM3062, DM3061, DM3058
Instek GDM-8255A
These are pretty unknown brands, it's hard to find any information about their quality and stability. Last time I looked, the Rigol bench meters were pretty close in price to the similar specced Agilent meters, in that case I'd choose Agilent.
Fluke 8808A
HP/Agilent 34401A
Fluke 8845A, 8846A, 8842A
I've heard some complaints about the split 2x4 wire bana jacks on the recent Fluke meters (8808A/8845A/8846A), they don't appear to survive very long (in professional environments). Apart from that, the 8845A/8846A are probably fine meters, with modern features like ethernet and USB (on the 8846A). The 8808A looks cheaper, only RS-232 for communication seems limited (no GPIB, ethernet or USB), and not great accuracy for a 5.5 digit meter.
The 34401A is a fine and popular meter, older than those Flukes (late eighties/early nineties?) and still in production. Tends to be more expensive used than many other bench meters.
The 8842A is significantly older than the other Flukes you list, from the eighties I think. It's quite close to the 8840A, I think the difference is better stability. AC and GPIB were optional. I described some alternative options from the same era
here. My favorite candidates (price/performance) would probably be the Keithley 196/199 (nice LED display), but all others on that list would also make fine bench meters. Note that a bench meter usually does DCV, ACV (sometimes optional), resistance (often also 4-wire), and often DC/AC current. Other features like frequency or continuity on bench meters are more recent developments. With used equipment, a lot depends on what you can get for a good price.
I paid €75 shipped and $25+shipping for two Keithley 199 meters (the former with returns allowed, the latter with some cosmetic damage like cracked display filter from a trusted source), so for me these were good deals, and $300-$400 for a used 34401A seems pretty steep. But I wouldn't pay
$110 + shipping for one old as-is that looks like a bomb went off in that room, or
$200 for one that looks good but is sold as-is, so based on what's currently on Ebay, I might choose something else. These Keithley meters have nice large LED displays, but the model 199 has weird current ranges (3A or 30mA, but the HP 3468/3478 only have 3A), and the max. voltage is 300V. This 300V limit was common for meters from that time, the same applies to the HP 3457A, 3468A, 3468B, 3478A and various other Keithley meters. The accuracy of the Keithley 199 is quite good for 5.5 digit, 90ppm 1 year basic DC voltage, significantly better than the HP 3468A or Fluke 8808A, quite close to the 6.5 digit model 196 actually.
Are you also considering the full-rack (as opposed to half-rack) instruments? They can be better value, eg.
HP 3456A for $150 (random ebay find), similar accuracy and resolution to the 34401A, but larger and less features. The equivalent Fluke offerings (850x) are also fine, but keep in mind that almost everything was optional, so get one with the right options. The best accuracy is always in full rack instruments, even now (eg. Fluke 8508A). Until the 34401A, most half-rack bench meters were limited to 5.5 digits.