I think it would be more reasonable to relate replanning to the distance traveled rather than time passed, or at least having behavior trees supporting both cases.

This would allow replanning rates to vary proportionally to speed and not replanning very short paths near goals that could cause bad boundary condition behaviors.