I have to wonder what was going through the minds of the Solartron engineers when they designed the 7061. While using the same specially selected 1N829A diode as the other meters that they made, they seem to have made no effort to keep air currents and temperature gradients away from the reference.
Here's the poor thing, flapping in the breeze at the rear right of the analogue PCB (D301A).
Unlike the 7075 (which also has the reference diode in an oven), which has air slots in the upper and lower covers, the 7061 is a sealed unit. In operation, there are large temperature differences within the case, mainly from the transformer, which heats up the left rear of the unit considerably. Here's what it looks like to my thermal camera.
All that heat is going to add up to some significant thermally induced air currents inside that case. With the unit tilted upwards on its stand, the air currents are probably even larger.
I have had the 7061 powered on since Friday evening, so it's now had more than a 48 hour warm up. Putting my Weston cell on the meter again showed good short term accuracy (+13μV compared to the calculated voltage), but I saw some +/- 3μV variations over time, which is something that my 7075 with its ovenised reference doesn't exhibit.
The solution is a simple as can be. I took a piece of cotton pad, split it down the middle and enclosed the reference diode inside it.
After letting the meter warm up again, I'm now seeing +/- 300nV variation on the Weston cell's voltage over a 30 minute period.