In this paper model, expiry = time obtained + time to live (TTL). Before expiry, use the cached address without restarting its lifetime. At or after expiry, the next lookup obtains the current source answer and starts its stated lifetime. The source is always available; no early refresh, serve-stale policy or page-content cache is modeled. A correct address does not guarantee a working page service.
Worked example
Separate example: a copy obtained at second 5 with TTL 20 expires at 25. A hit at 21 leaves 4 seconds; it does not move expiry to 41. A lookup at 25 refreshes from the source.
Observatory A / fictional lookup log
At model second 12, the resolver caches atlas.test = 198.51.100.41 (Mirror), with TTL 48 seconds. At second 36, the authoritative source changes to 198.51.100.40 (Atlas). Requests occur at seconds 51, 59 and 60; there are no other lookups. Every refreshed answer has TTL 48. Routing to both services works. Atlas page service is stopped at second 60. Times are seconds on the same model clock. Do not contact or configure these documentation-only addresses.