You have not provided enough details. But, I am willing to take a guess:

The cyclone separator is undersized
The particles are lighter than water
There is not enough differential pressure between the flush supply and the cyclone drain
The orifices into our out of the cyclone are eroded and oversized

Based on our limited description of the service, I am picturing a high speed Sundyne in wash water service. The cyclone separator could be internal to the pump case. These are notorious for plugging up or eroding away. The orifice that feeds the separator may be the style that uses a little o-ring to seal the fitting to the inside of the diffuser. This o-ring could be missing or not sealing, allowing flow to bypass the orifice.

Provide a proper description of the pump and the service and we won't have to guess.

Johnny Pellin

Much more information is needed. We have seen instance of corrosion from the end user not understanding the metallurgy required in the RO with the fluid and the particulate together. This can cause severe degradation over short periods of time. How often are you having a shutdown with replacement and inspection?
Mr. D

aasb1988 - you already determined why the seal failed. Due to particulate causing damage on the seal face. You need to determine where that particulate is coming from, whether or not something was supposed to have removed it (most likely for this reason as well as others), or whether something is generating it which shouldn't. Sometimes unplanned corrosion happens from material incompatibility, or water chemistry issues, or mistakes in fabrication, or a mistake in filter size - but whatever the reason or failure we have no way to determine that with the information presented.

Quote (jjpellin)

The cyclone separator is undersized
The particles are lighter than water
There is not enough differential pressure between the flush supply and the cyclone drain
The orifices into our out of the cyclone are eroded and oversized

Add to that list my favorite - haven't cleaned the filter, so the cyclone is plugged and not separating (I supposed lack of ΔP might be the same thing).