[MMinasyan]

Summary

Share repeated Harness and Runtime test construction while retaining scenario inputs and independent observations.

Changes

• Reuse pending-call graphs, raw codec rejection rows and idle-source Fork setup without removing corruption or ownership checks.
• Share scripted execution values and admitted-session setup while preserving callback roles, cancellation, lifecycle gates and Close observations.
• Group equivalent successful provider key-source and disconnect cases with isolated runtimes; retain the external-key refusal and its baseline separately. Reuse metadata readers, interrupted-tool fixtures and fixed session-read callbacks.
• Keep private/public, memory/SQLite, raw-wire/generated-client and pure/live observations distinct. Remove 269 net test lines without changing production code.

Verification

• Named-case and observation maps preserve the original scenarios through their existing boundaries.
• Representative graph, model-role and provider-key-source mutations fail against both original and consolidated construction, then are restored.
• Each commit passes a Wails build and binding check; separate full Go and race suites; vet, formatting and lint; full frontend tests (510 passed, one skipped); protocol freshness/typecheck; installer tests.

Limits

No production behavior change, backend removal, cross-package test migration, new fixture framework or client-side consolidation.