I know this is a while ago now, but maybe helpful to others...try using the "hidden" dimension `_timeseries`. This is a JSON string that is an amalgamation of all of the dimensions for each datapoint. Take care, the results may be (very) high arity and splunkd doesn't (yet?) have very strong protections for itself (in terms of RAM used while searching) when using this code path, so it is (IMHO) easy to crush your indexer tier's memory and cause lots of thrashing.
... View more