Currently the Accumulo optimization to flatten fields sharing a common visibility attribute into a single column qualifier serializes/deserializes the original field IDs alongside the values. This approach increases the size required to store each feature and we also pay a performance penalty.

Investigate approaches to avoid the serialization/deserialization of field IDs while maintaining the ability to properly deconstruct the flattened value (preferably without knowledge of the data adapter) and support field level query filtering.

Ultimately this should result in a performance boost, but the tradeoff is additional complexity. Goal is to find the best approach to achieve this performance boost while minimizing complexity.