The survival curve
One cohort of — coins, launched between — and — ago, followed forward through its own history. For each, every swap it ever had — so a lifespan is simply the time from launch to last trade.
This follows one set of coins forward, so no threshold enters the definition and no cross-sectional caveat applies. The point at launch is the share that outlived its own creation block — a little lower than the share that ever traded, because — coins had every one of their trades inside that first block. Coins still trading when the chain was scanned (— of the cohort) are counted at the lifespan observed so far, which can only pull the median down, never up. Each point divides by the coins old enough to be eligible at that age, not by the whole cohort.
How firmly is the median known? On this scan the 95% interval runs —, from order statistics — both ends are real lifespans out of the cohort rather than a modelled curve. Across scans the figure has landed anywhere from 2.3 to 3.4 minutes, so treat it as "a few minutes" rather than a stopwatch reading. It will tighten as history accumulates and successive cohorts can be pooled.
Dying gets slower
Each successive quarter of the cohort takes several times longer to go than the one before. Surviving is evidence of being the kind of coin that survives — which is why the expected remaining life of a launch rises as it ages.
How much longer?
Of the coins that reached a given age, the median time they went on to last. Straight from the cohort, no model fitted — and a statement about the cohort's dead coins from a past six hours, not a prediction about any particular one. Nothing here says what any coin will do next.