An exploration node can keep producing goals long after useful map growth has stopped. A credible system needs more than a loop that sends the robot toward unknown space.
Bounded exploration
Frontier candidates should be filtered by reachability and information value. Repeatedly failing the same goal needs a blacklist or cooldown. The system needs stop conditions for low frontier gain, exhausted reachable candidates, time or distance limits, and operator cancellation.
Simulation can reveal oscillation, unreachable frontiers and poor goal spacing, but physical deployment still adds sensor, traction and controller uncertainty.
The useful outcome is not continuous motion. It is a map whose coverage and remaining unknowns can be explained.