When a near cache is used it will install a remote listener that receives all modified/remove/expired events. If a client is unable to keep up with the amount of updates, this can cause a backup on the server which in turn can slow down writes.

One solution for this would be if a client has been backed up for a given threshold to remove the pending events and send a single clear event to the near cache. This would take away the read performance temporarily of that client temporarily but would allow write heavy clients to continue operating.