IIRC, last time I played with an ICL7660, the shoot-through was quite substantial both on the input and output sides. And it's painfully slow, so the shoot-through and ripple puts considerable energy below 10MHz, which is hard to filter: ferrite beads are ineffective.
On the upside, most op-amps have acceptable PSRR in that range, so it shouldn't be too big a problem. If it is, your first solutions should be:
1. Ditch the 7660. There are faster charge pumps, and much faster inductive switchers -- easier to filter, and some of them will be quieter to begin with (as strange as that sounds, but yes, the 7660 IS just that awful).
2. Cut out the bad-PSRR parts. (RF sensitivity is rarely documented, unfortunately.) Choose better op-amps, isolate the sensitive stages, consider local filtering, etc.
3. Consider adding LDOs, or even better, C multipliers.
The tricky range is kind of 10kHz to 1MHz. Passive filtering is bulky, switchers are noisy, and LDOs and op-amps perform poorly. Pushing the noise above this range (using a switcher in the >300kHz range, say) makes it easier to filter; pushing it lower (not really an option for inductive switchers, but I suppose you could put electrolytics in a charge pump running this slow) makes it less of an issue, but only if the harmonics don't screw you up in the process (which is where the '7660 is at, unfortunately).
Tim