[FLO-13.6] Prove the temporal release workflow with released providers #33

Parent: #13
Depends on: #28, #30, #31, and the real multi-provider foundation from #3
Provider evidence: aniflow#29, aniflow#24, renderflow#412, optiflow#65

Outcome

Prove one clean-room time-based release workflow through independently built or immutable Aniflow and Renderflow artifacts, including Optiflow only where inspection of poster, frame, or output assets is semantically appropriate:

immutable synthetic temporal source → Aniflow processing/analysis → Renderflow derivatives → optional Optiflow inspection → validation → aggregate release evidence

Dependency boundary

Reuse the completed Aniflow provider-conformance foundation from #29 and the smallest accepted immutable fixture subsets from Aniflow #24 and the other provider corpora. Full corpus completion is not required if the precise released fixture artifacts and contracts needed here are available.

Scope

• Pin exact provider artifact, contract, codec/tool, and fixture identities.
• Preserve timing, stream, frame, artifact, and provenance evidence across provider boundaries.
• Exercise one positive multi-provider workflow.
• Cover unavailable capability, partial/invalid output, corrupt derivative, insufficient validation, and provider-version mismatch.
• Interrupt after a completed temporal stage and prove deterministic resume.
• Prove timing, source, configuration, or provider changes invalidate the correct dependent work.
• Keep optional Optiflow branches honest when no semantically useful inspection applies.

Acceptance criteria

• The workflow runs from immutable provider artifacts without sibling source imports.
• Each provider remains independently executable.
• Timing and artifact invariants are validated rather than inferred from exit status.
• Interruption and resume reuse accepted work without duplicate temporal processing.
• Optional branches report skipped or unavailable honestly.
• Tests require no proprietary media, network service, paid API, GPU, or model download in the ordinary tier.
• Expected manifests and normalized evidence are reviewable, deterministic, and budgeted.

Non-goals

• Implementing codecs, temporal algorithms, rendering logic, or optimization logic in Flow.
• Requiring probabilistic analysis as ground truth.
• Publishing or uploading real media.