[pelikhan]

Why

The core Go test job spent 9m35s running tests serially. Sixteen top-level scenarios taking at least 10 seconds each accounted for 4m28s of that execution time.

Changes

• Partition core packages into four deterministic, disjoint shards while retaining storage test-name sharding.
• Move those 16 slow CLI-build, fleet simulation, PostgreSQL integration, conformance, and benchmark scenarios out of PR and main-push runs using testing.Short(). Daily and manual runs retain full coverage, including the service-backed dashboard suite.
• Extend the existing daily full-validation workflow to create or update one bot-owned failure issue with the latest revision, suite outcomes, and run/job links. Issue writes are restricted to the main-branch reporting job. Successful runs leave investigation issues open for human triage.

This intentionally trades exhaustive per-push coverage for faster feedback; the existing 06:37 UTC daily run remains the full-suite backstop. The 4m28s figure is historical test execution moved off the frequent path, not a measured end-to-end speedup for the new CI jobs.

Validation

• 37 focused CI contract tests passed, including shard completeness, deterministic selection, test modes, issue deduplication, and API error propagation.
• All 16 selected scenarios passed against disposable PostgreSQL in full mode and skipped in short mode.
• Actionlint, ESLint, Go vet/format checks, CAO and Copilot type checks, and workflow compilation passed.