Common-acceleration complete orientation comparison

Repeat all six original generated cases and all eight original yaw sectors with the qualified common-acceleration construction. Keep byte-identical 400-packet inputs, the original 27 physical bounds, six training times 0.2/0.35/0.6/0.9/1.2/1.5 seconds, future time 1.8 seconds, 1 mm training allowance and 10 micrometre arithmetic allowance. Retain every physical and all 15 shared error columns with a single fixed source/domain/expansion binding per sector.

The 24-identity native and 19-identity joint-interface qualifications must be complete, passing, hash-verified and in the same numerical environment. Contract only necessary training inequalities. Reuse the fixed physical expansion, bound the containing hull when necessary, and preserve every nonexcluded sector in the reported union. Save all native values/Jacobians, reference definitions, inequalities, common blocks, remainders and certified support inputs before evaluating future truth. The known-control common vector is only a subsequent retention oracle.

Correctness requires native/bound centers within 1 micrometre, retention of each true physical-plus-common-error point in its original sector, and enclosure of its known future output. Preserve the original tenfold usefulness target against the broad-box baseline on every axis in all six complete cases; additionally report change from the completed shared-reference comparison. Never select sectors or change bounds using future truth. No original recordings, observed fits or protected ESA outcomes are used.

Limits: 48 identities / 144 native calls / 57,600 native packet visits, 384 bounds / 153,600 reserved bound steps, 48 contractions, 288 LP proposals / 576 support certificates and a separately frozen 3,600-second cooperative ceiling. One worker reserves four numerical threads. This candidate's exact matrix-norm arithmetic makes it slower than the earlier construction; the 122-second native qualification motivates a longer finite ceiling. This does not expand any completed allowance. Charge/reserve before operations; no retries, expansions or process signals. Preserve failures and terminal outcomes.

A short relative-width improvement is not recording-scale usefulness, a full floating-point certificate or calibrated observational coverage. Those remain separate gates before new aircraft prediction work; all scientific outputs abstain.

Native qualification · Joint interface · Previous comparison · Roadmap.