1 box · reads per 15 minutes
Grouped by box, not by mat, because a box is what you drive out and fix. One box often covers several splits, so its failure shows up in several places at once. The bars are reads per 15 minutes across the race, coloured by distance, on one shared scale so boxes are comparable.
How many athletes a mat read, out of how many we can prove went past it. An athlete counts as proven only if a later split recorded them, or they finished. Somebody who vanished from the results entirely is never counted against a mat, because we cannot show they were ever there. That makes every number here a floor on real losses, never an overstatement: the true miss count can be higher than what is shown, never lower.
One physical timing unit, identified by the box id the timing system assigns it. A box is what you drive out to and fix, which is why these pages group by box rather than by mat. One box commonly covers several splits at once (an aid station read on the way out and again on the way back, or a mat shared by two distances), so when it fails, it fails in several places simultaneously.
Each one names a different repair, which is the whole reason they are separate rather than one "unhealthy" label.
A verdict always describes a box's worst mat, not its average across mats. A box reading 99% outbound and 10% inbound has a serious problem at one of them, and averaging to 55% would describe a box that does not exist.
One bar per time bucket, coloured by race distance, with empty buckets kept so silence is visible. All cards share one time axis and one height scale, so bars are directly comparable between boxes. That scale ignores the single busiest bucket, because every runner crosses the start within a bucket or two and scaling to that spike flattens the rest of the day; a bar that runs past the scale is capped and marked, and hovering always reports the true count.