I have successfully used tetraspek beads as fiducials for STORM/PALM. This was a long time ago and I can't remember if the detector (Andor iXon DU-897) was saturated at those pixels, but in either case it worked fine. Some people use gold particles which fluoresce across wavelengths. I found gold to be very dim in my hands and never found a good way to evenly disperse across the coverslip.

Gold nanoparticles around 100 nm diameter work ok; a little brighter than mEos2/3.2 with 561 nm excitation and typical RFP-ish emission filters. In my experience, something a bit brighter than that would be even better, but they work well enough if you have a few particles to average and/or have a moving average position over a few frames. Also because they aren't super bright they need to be in good focus. I would like to try >=200 nm diameter next time.

If you happen to be working with a fixed sample, an alternative is not using anything and using the data itself as drift correction; works if there is only slow xy drift.