Up to know, flagd has not handled flags with conflicting keys from different sources, or even the same source.

As described in these 2 documents, we want to support a sort of "flag namespacing" via our existing flagSetId property. Logically, the unique identifier for a flag would be represented by a combination of flagSetId and key, irrespective of the sync-source.

Definition of done:

• flags are uniquely identified by flagSetId and key in all endpoints/protocols (evaluation.proto, sync.proto, OFREP); clients identify their flagSet (either at steam initialization or during the RPC, depending on the ptotocol) and then get responses according to that flagSet.
• flag metadata and flag set metadata are restricted similarly to above
• the means of expressing the flagSetId must be extensible, so that other criteria and filters might be supported later