There is a figure 52 on the AD8622/8624 datasheet that shows the crosstalk on the dual.
For any part from any manufacturer, it is important to know how it was tested. If you consider
one of the sections to be the transmitter, and another section to be the receiver, then you want
a heavy load on the transmitter, and no load on the receiver. With what we have learned about
layout in the past 30-40 years, the duals and quads today are pretty good. Because of this,
the interference from the transmitter is very small, so you want to take gain in the receiver.
Note that figure 52 uses a 10k load on the transmitter. The AD8622/8624 is a low power part,
so the spec tables have Voh and Vol specified with 10k and 100k loads. Thus, fig 52 uses
a 10 load. The gain on the receiver is +101. When we graph the data, we take this into account,
so the dB scale is input refered.
The low frequency crosstalk is a thermal effect, the power dissipation in the output stage of the
transmitter travels across the die and unbalances the input stage on the receiver. The high frequency
cross talk is generally a capacitive effect, both in the test setup, but also the capacitance to substrate
for all the transistors in a junction-isolated part.
Another disadvange of using quads, especially with active filters, is you wind up with 20 passives
crowded around a tiny package, and the pc board trace to trace capacitance is a killer.
If you need extremely good numbers, then using duals or even singles would be the way to go.
See the article(s) below for more details.