> how much the shift setting matters here A lot, because website fingerprinting papers can look great on one trace split and then fall apart when the timing, batching, or transport behavior changes, so I'd want to know exactly which shifts they tested before trusting the robustness claim.