HP 3457A teardown, some poblems + needs full calibration

[Kleinstein]

The slight negative rail can be desirable for the 300 mV and 30 mV case. There is no need for it with the 30 V circuit.
The diode for the normal ground is the simple way to a small negative voltage, provided there is more load from the posivide side to ground that from the negative side. So OPs that don'e need the neg. side should still use GND for there neg supply.

The MLC caps of X7R or similar type can be a slight problem at the ICL7650. X7R caps show quite some DA that acts a bit like leakage for the first few hours of use.  It may still work, but may need many hours, maybe days to settle.
I don't think the caps are fake, just the wrong type. There normal choice are polyester caps like MKS series. These are not really expensive either (usually some 10-30 cents each when you buy 10). The external caps are also a bit prone to react to some hum pic-up on the breadboard. So the AZ OPs with external caps went a bit out of favor, though they still have a few good points (relatively low bias / current noise).

The OPA991 is not an extra precision part, though one can be lucky and get a low TC one. It is kind of OK for the 3 V / 30 V level. The divider from 3 V to 300 mV could be measured with the DMM directly too.