An engineer deletes the only static route to a remote substation subnet entirely — not replacing it with a different route, just removing it — while a technician has an active session to a device in that subnet. What is the most accurate expectation for that already-established session?

While an established session's forwarding decision is cached at session creation and is not re-evaluated packet by packet against every routing table change, that caching does not make a session immune to routing entirely — once the only route to a destination is deleted outright, with nothing left standing in for it, the session's traffic ultimately has no valid path to continue over, and the session fails, whether through interface-level effects, expiration, or the next event that forces a fresh route lookup. This nuance separates a route being changed to a different valid alternative, where the existing session can persist unaffected, from a route being removed with no replacement, which removes the reachability that traffic to that destination depends on. Claiming a session is unaffected forever overstates the caching behavior into something FortiOS does not actually guarantee. The FortiGate does not automatically recreate a route an administrator deliberately deleted; that would defeat the purpose of the deletion. Sessions also do not spontaneously convert themselves into policy-route-governed traffic; policy routes are a distinct, explicitly configured feature, not a fallback a session adopts on its own. A practical check is to delete a test route to an otherwise-idle test destination and confirm that an active session to it eventually stops passing traffic rather than continuing indefinitely.