Constraints and viable futures
Can targeted restrictions steer horizon-limited decisions away from states with fewer viable future targets?
RP005A reviewed and completed
24 planner rows, 12 paired comparisons and 35 structural sources independently replayed. Two positive, four zero and six negative outcomes remain preserved: V1 benefits at H=1/2, with no benefit at H=4; V2 and C0 lose coverage. Pure edge removal never expands structural reachability. D-0007 closes the finite RP005A commitment after internal package review. Q-2 stays ACTIVE; Cycle 01 is PUBLISHED. 13 / 13 obligations resolved (100.0%).
Structural reachability versus a planner choice
Pure edge removal cannot create a reachable node from a fixed source. A supplied restriction may change a truncated reward maximizer's first action. Viability here means that a declared terminal target is reachable within three future transitions; it is not indefinite safety or realized visitation.
Comparison contract
Greedy H=1, short H=2 and sufficient-for-these-fixtures H=4 use exact integer reward maximization with lexical destination tie-breaking. Execute one action. The fixed five-target denominator stays the same in both arms. Primary coverage uses each arm's admissible graph; secondary evaluates both selected endpoints in the original graph. Their decomposition separates selection gains from deleted future paths.
V0 removes an unused branch. V1 blocks an attractive delayed-cost branch; a sufficient horizon already avoids it. V2 blocks a useful branch. C0 deletes a downstream target path and catches apparent gains from changing the denominator. All expected outcomes are disclosed engineering controls, not independent discoveries.
Evidence and package disposition
The single recorded audit completed under its sixteen-input freeze; all planner values, target sets, pairs and structural certificates independently replayed. TST-RP005A001 has scoped Result R-RP005A001 (VALID). Selection and deletion terms reconstruct each primary difference exactly; negative and zero outcomes remain visible. D-0007 accepts the finite commitment after explicit internal package review. RP005A is CLOSED; Q-2 remains ACTIVE. Cycle 01 is PUBLISHED; 13 / 13 obligations resolved (100.0%). D-0008 completes the late checkpoint and graph review; D-0009 completes synthesis and foundation reviews. D-0010 completes change/open records. D-0011 accepts the report-scoped legacy pilot and transition. D-0012 subsequently accepts the final close bundle and canonical snapshot. The full-information restriction designer, tiny graphs and reward/target mismatch limit interpretation.