AD8426 driving adc goes unstable?

Datasheet is very clear on using AD8426 to drive an ADC. It states clearly that a capacitance is needed to supply current to the ADC during sampling and that a resistor is needed to shield the AD8426 from this capacitor to keep the amplifier stable. It describes necessary time constant as above 5µs.

The problem is, extra capacitance in the system makes the AD8426 unstable while any extra capacitance only serves to increase the time constant. In fact, a larger R or larger C will both increase the time constant. Always preserving the requirement while real world experience is instability.

Is there an error in the datasheet?
Perhaps a miswording?

The suggested 100Ohm for 100nF delivers 10µs time constant and 15915 Hz frequency response. Just wondering if something is amiss.
Need to figure the limitations before we go into production...