The maximum achievable performance of strange-jet tagging at hadron colliders and the loss in performance in different detector designs is estimated based on simulated truth jets from strange-quark and down-quark hadronisation. Both jet types are classified with a recurrent neural network using long short-term memory units, at first using all available truth particles and then applying selections to study the impacts of ideal tracking detectors, Cherenkov detectors, and calorimeters. Additionally, a manual reconstruction of strange hadron decays such as KS → π+π- from charged tracks is considered.