Hi Jim
You are right that it is usually not an issue but it is worth knowning about possible limitations. I remember using the 5213 in a project which was measuring NTCs, which can have resistances of about 100k and we did use op-amps in that case. However there was also some active filtering too, although I think that a large capacitor would probably have been as good (slow signals - and the large capacitor [large in this case would only need to be a few nF] also acts as a good current source for short periods).
I am no longer sure of the details but I think that the sampling period (the time that the input is actually connected to the sample-hold) is about 200ns at 5MHz (maximum ADC clock). A resistive source of 30k would take about 200ns to charge the 1.4pF input capacitor to the input signal level (using 5xRC as approximately fully charged). Any stray capacitance or real capacitors in parallel help charge, although they can have a filtering effect on the input signal too.
Therefore I suspect that any source with less than 10k should not be an issue at all.
Probably using a slower ADC clock also increases the connection time.
Regards
Mark
www.uTasker.com