How long evidence from a live feed stays current
A live estate publishes its boundary feedBoundary feedRecords supplied by the customer’s own systems about activity at the edge of an agent or service. NovusLens reads those records; it does not watch the agent. with a refresh interval. When NovusLens reads that feed, every findingFindingA difference in the records that needs a decision. It includes its sources and can describe a slow handoff, disputed fact or missing owner. built on it is supported for three of those intervals after the entry was written. The acquire command prints the time its support runs to.
After that time the feed has not been read again, so its evidence is no longer current. The findings it supported are withheld. Their identity and the reason (feed support lapsed) stay visible. Their claim is not repeated, and nothing is deleted. Reading the feed again restores their support when the refreshed evidence still supports the finding; a finding the fresh feed no longer supports stays withheld.
This is the same rule for every feed, simulated or real. A one-off read is a demonstration; a live estate is read on a schedule the operator declares and can see. A notice recipient who was already told about a finding hears when its support lapses; the finding itself is not delivered again.
What to do: put the acquire on the agreed schedule, and before relying on a boundary finding check the time its support runs to.