As the CSV cache is a write though host side cache, the only effect should be to reduce read requests to the array. This buys you more read IOPS to use for something else. Writes will be unaffected.
The only minor effect on the array is that blocks might be moved out of the Nimble cache as they are accessed less frequently. The only time I can think of when this could make a difference is when moving services between nodes in the Windows cluster as data cached on one node probably won't be cached on another. You might have a few extra disk reads on the array when this happens which will repopulate both the CSV cache and the array cache.
Overall I would expect the frequent advantage of host side caching to outweigh the infrequent disadvantage of data not being in the array cache.