Registered Strategy B on the T-S02 constant-objective profile

Single-profile exploratory evidence. One excluded warmup and seven saved observations; no comparator or speedup claim.

Solver source e7f66a9ee698d05ed8779621116319158fa70158; separately versioned harness cabd2f37199e2f8412d75328cf7b881e590a468e. Input SHA-256 b080fb248768be037f35e1128a6533a89e2f3b1695292ff2b839c89f4829b9a3.

Assumptions

Discrete two-step scalar integrator at times 0, 1, 2.
Known initial state x0=0; all state constraints are universal.
Both transitions are x[t+1]=x[t]+u[t] with A=B=1, zero offsets and no external channels.
Continuous controls independently constrained to [0,1].
Nominal open-loop information; no observations or temporal constraints.
Objective identically J(x2)=3, represented by LinearTerminal with coefficient zero and offset 3.
Exact optimal joint-history H-relation requested under Assurance::numerical.

Full optimal joint-history relation

Every feasible history has objective 3. The complete relation is (x0,x1,x2,u0,u1)=(0,s,s+t,s,t) with 0≤s,t≤1. Each timed run checked completed optimal status, registry-verified exact family, objective bounds [3,3], and every coefficient of the canonical H-relation.

0 ≤ u0,u1 ≤ 1; x0=0; x1−x0−u0=0; x2−x1−u1=0

The distinct histories (0,0,1,0,1) and (0,1,1,1,0) both have terminal state 1. Each timed run checked both, the endpoint histories with terminal states 0 and 2, and two infeasible nonmembers. All membership arithmetic uses exactly representable binary64 integer literals.

The registered completeness verifier is csl.verifier.ts02-constant-integrator@1, with Assurance::numerical. The production analytic branch constructs the full feasible relation. It prepares the registered production providers but invokes no LP/QP solve; this report does not measure native solver latency.

Observed public-call timings

Fresh problem construction and resolution, registered Strategy B and production HiGHS LP/convex provider creation/selection, StrategyRegistry preparation, exclusive public StrategyPort workspace creation and run, compiled T-S02 verifier acceptance, objective bounds [3,3], literal canonical full joint-history H-relation and membership checks including two distinct histories with equal terminal state. Includes local owner destruction before one_run returns. Excludes process startup, JSON serialization, build, file persistence and rendering.

Seven observations, median 57200 ns; min–max 51900–74200 ns; range width 22300 ns. Failures: 0. The excluded warmup is retained in JSONL/CSV.

Each dot is one saved observation; no outlier was removed.

Repetition	Elapsed (ns)	Objective bounds	Result disposition	Validation
1	74200	[3,3]	complete_optimal_family	passed
2	59800	[3,3]	complete_optimal_family	passed
3	59900	[3,3]	complete_optimal_family	passed
4	57200	[3,3]	complete_optimal_family	passed
5	52300	[3,3]	complete_optimal_family	passed
6	51900	[3,3]	complete_optimal_family	passed
7	54600	[3,3]	complete_optimal_family	passed

Performance conclusion: inconclusive. The separate diagnostic workload and process counters do not establish a source-mapped limiter on the unprofiled tiny analytic T-S02 path. Performance remains inconclusive; no latency limiter is claimed. The separate diagnostic run is retained in profile evidence; its durations are excluded from the seven observations.

Evidence limits

Completeness applies only to the exact admitted T-S02 constant-integrator profile and csl.verifier.ts02-constant-integrator v1.
Assurance::numerical is reported; validated arithmetic is unavailable.
Seven observations of one tiny profile are exploratory evidence only; no comparator, speedup, general performance, application-scale, throughput or scalability claim.
This report does not cover T-S03, singular A, other dynamics or objectives, or generic complete-family support; SYN-04 #67 remains open.
The admitted T-S02 production path constructs and verifies the full feasible relation analytically; LP and convex providers are registered and prepared but their numerical solve operations are not invoked.
The separate diagnostic workload is excluded from the seven observations; no source-mapped unprofiled latency limiter is established.
Internal preparation, native LP/QP, relation construction, compiled verifier and export timings/counters are unavailable.

Environment and reproduction

Host/compiler/Release settings, exact source and harness identities, dependency hashes, executable hash, clean-source checks, pre/post host load, resources, warm-state policy, commands, and unavailable phase counters are retained below. Absolute host roots are replaced by portable placeholders. The build provenance retains original evidence hashes separately from the hashes of the sanitized published files. The interval includes registry preparation and verification; it is not a native-only solve interval.

Typed JSON · Raw JSONL · Raw CSV · Measurement · Scenario · Environment · Build provenance · Commands · Separate profile · Reproduction notes

This partial SYN-04 evidence keeps #67 open. Deployment and live URL verification are separate main-only gates.

Publication syn04-strategy-b-ts02-3999b008c529fd57b5cfb6a78841a9963a7e9c4c9919e0c4edafe705b236f929; source report syn04-strategy-b-ts02-20261014-e7f66a9e. Versioned envelope · Publication manifest