I tested the accuracy of my 34465A:

I used a 1kHz signal from the MSO5104 signal generator at 600, 300, 150, 100, 75 and 50mV Vpp output. I measured the RMS using two methods - using the Vac measurement from the Keysight, and the digitizer. For Vac I used the graph over time to get an average value. For the digitizer I took between 6 - 12 sine waves, measured Vpp and calculated Vrms. I also measured the input to the resistor divider I built (using 100kΩ and 1kΩ 0.1% metal film, 5ppm/C Ohmite resistors) using the Keysight Vac measurement, and previously I calculated the average actual resistance of the two resistors over a 20 minute period. Using ohm's law I calculated the theoretical value.To me it seems like the digitizer is pretty good down to 174µV - with an error of 1.38%. The Vac function started diverging at approx. 350µV with an error of 7.57%.

Only issue is the digitizer only works for pure sine waves - I cannot easily calculate RMS for mixed frequency signals.