We're currently using Coherence 126.96.36.199 and we're running into the same situation here, we have some use cases that require heavy querying on (typically) non existent keys on a near cache. It would be very useful for us to have the ability of setting some cache-miss configuration to prevent them to go through the back cache.
For the time being we worked around this query overflow on the distributed cache by setting up a runtime cache at the application level which is updated through listeners on the distributed cache (i.e. we're using it as a front cache at the application level), but we'd like to come up with a full solution when something like this is available. How is this internal bug going?