How long disruptions take to recover

Across 20 ended events with a measured, material fall, 18 had their series return to at least 90% of the pre-event level by 2026-09-13 (13 September 2026); 2 had not. The median recovery time was 317 days, a figure that mixes very different kinds of events: a short storm closure back to normal in under two weeks alongside a bankruptcy or a years-long capacity shift where the pre-event level was never really the point of reference again. This counts days from the end of the measured event window to the first 7 day trailing average at or above 90% of the baseline: a plain reading of the same public series the effect method uses, not a method figure, and not a claim that recovery was caused by anything in particular. A row only counts as a fall here when it also clears the severity rubric's weakest bar for a real one, a standardised score at or below -2: without that, a small, noisy dip can drift for years before a trailing average happens to cross the line, which reads as an implausibly slow recovery when nothing measurable happened.