Finite acceptance map

All assertions are source-bound; expected roots, offsets, schedules and tree shape precede the implementation. The initial 17 materialized seed carriers total 107086 bytes. Sixteen are valid finite carriers; the last is a synthetic maximum-depth prefix, not a witnessed complete huge file. Malformed runtime variants are explicit mutations in the full test source, not separately materialized in that seed count.

• proof.rs: literal schedules, whole-group coordinates, request intersection, pointer identity, two independent cursors, terminal output, external quotas, total exhaustion, cancellation, interleaved jobs, view/cursor retention and Debug.
• proof_failures.rs: geometry/error precedence, all canonical-word positions, short headers, every truncated prefix of selected carriers, parent/leaf errors, strict suffix rejection, terminal stickiness, grouped ROOT and absolute counters, all BlockSize logs, deepest finite prefix and largest-width setup, exact Display.
• proof_allocation.rs: observer positive control, successful verification/output, terminal polls and cancellation; EOF failure and constructor geometry failure. Other error branches are verified functionally, without an allocator observation claim. Fixture allocation and destruction happen outside the observed interval.
• seals/src/bin/*: positive runtime control plus private fields E0451, private constructor E0624, raw conversion E0277, DeserializeOwned and Deserialize<'static> E0277, proof drop E0505, proof mutation E0502 and immutable leaf mutation E0594. Exact compiler classes are asserted; generic compile failure alone is insufficient.
• interleaved-oracle/: independently assembled four-leaf carrier from committed kernels, never a traversal/encoder/verifier. Its parent pair after two payloads exercises resumed frontier state and replay offsets. Original adapter manifest's 73 output actions includes its planned copy work; Radio emits six iterator items.

Negative compiler API absence on b933 is recorded separately from actual legacy carrier runtime observations on a96d. Original tests-first and intermediate source freezes are preserved. The final tests-only interleaved witness has its own source bridge and focused debug/release results; earlier broad counts keep their 17-test source epoch and are not relabeled as runs of the expanded corpus.