Track A prospective. Retain exact source size and the observed odd-count label. Search for positive-prefix coalescence at the same depth and same total odd count, beyond yesterday's length-31 range certificate. Enumerate actual odd sources in [1,2^24), CPU int64, at most 96 shortcut steps and prune nonpositive coefficient prefixes. Search duplicate (endpoint,odd count) within each depth. Guard overflow before multiplication and fall back to scalar Python for exceptional large states. Save the first examples and independently replay them. Stop after 180 seconds or all 96 depths. If found, quantify the minimal span and its affine correction. Absence in this finite range is not an injectivity result.
Track B prospective. Use the cancellation-preserving grouping of Wren Paper115 on the exact future-success ledger. Compare full prefix data, (height,quota,last-bit), and quota-free grouping. Prove the exact minimum sufficient state for the fair forecast and a finite cap-to-infinity correction. Test the purported savings on exact integer cohorts of bit lengths 10,14,18,20 and at most 32 future bits. Bound all source loops and exact DP. A generic regrouping identity is background; a publishable result needs a new usable inequality, an exact obstruction or a quantitative certified consumer.

