We have noticed many flows are not enriched in NDH scenario. This is likely due to some latency in FLP receiving updates from the kube API, as NDH is spawning many pods that would quickly run requests against a database.

The goal is not to solve definitely this latency issue, it's certainly unavoidable if we want to keep our own collection process fast & efficient, however we may investigate and see if playing with FLP's kube client QPS/Burst would help decrease the quantity of unenriched flows.