[Proxicon]

Closes #455

Follow-up to #452 and owner acceptance of v0.1.3-beta.3.

Root cause and reliability

• Compare collection timestamps at PostgreSQL microsecond precision across scoped upload requests; keep the original wire hash so existing full-precision durable outboxes and duplicate acknowledgements remain compatible.
• Preserve complete-snapshot reconciliation, tenant scoping, conflict rejection and late/out-of-order protections.
• Bound accepted scans, surface effective timeout states on read without read-side scheduling, allow explicit retry and retain the previous complete snapshot.
• Bound agent requests/collection, keep failure backoff, and reduce healthy work-discovery delay.

Operator UX

• Load saved rows independently from status/refresh requests; reconnect status polling and fence cancelled tenant/search responses.
• Reusable full-screen, single-scroll client workspace with readable status, received-chunk progress, stalled/error explanations, metadata details and explicit refresh.
• Estate overview cards and Software / Installations / Devices / Changes views, searchable client-name picker, paged MudBlazor tables, drill-down and descriptive empty states.
• Resolve friendly names / reported hostnames server-side for UI, CSV and existing MCP/common-client responses. Include never-scanned devices rather than treating them as empty installation search results.
• Add tenant-authorized overview and device-coverage REST reads. Existing managed MCP operations retain their names and admission behavior.

Focused regression coverage

• Real PostgreSQL, 205 packages, fresh DbContext for each chunk, deliberate sub-microsecond wire timestamp, duplicate replay and genuine metadata-conflict rejection.
• PostgreSQL coverage/hostname/device queries and tenant isolation.
• UI saved rows visible while status or refresh POST is unresolved; first-collection state and existing tenant-switch cancellation.
• Expired-scan recovery retaining the last complete snapshot.
• Updated canonical software inventory guide.

Validation and release scope

Validation will be recorded from the existing GitHub-hosted PR workflow; no local release builds or fabricated native acceptance results.

No version bump, tag, release publication, container publication or production deployment. Merge to main only; owner is collecting further fixes before the next beta. Live native Windows/Linux and visual deployment acceptance remain owner-run after that future release.

[Proxicon]

Validation completed

Validated head a6eb22f0297f4897b2520c3cf034f86642c5dc50 against main at 9103886b792b62af7851d7d027e7ba6e843f50ed in Public PR validation run 41307104266.

All required jobs passed: Public disclosure gate, Source and release contracts, .NET build and tests, and the final PR validation gate.

• Release solution build: passed, 0 errors (compiler/analyzer warnings remain).
• Web component tests: 670 passed, 0 failed, 0 skipped.
• Core tests: 4,724 passed, 0 failed, 0 skipped.
• API integration tests: 36 passed, 0 failed, 0 skipped.
• Total selected fast regression suites: 5,430 passed. All three TRX receipts were validated; no tests or CI gates were disabled.

Coverage includes the real PostgreSQL 205-package / three-chunk upload with a fresh DbContext per request, full-precision queued-payload compatibility, duplicate replay and genuine metadata-conflict rejection; cached rows rendering while status/refresh is unresolved; first-scan state; tenant authorization; hostname resolution; and timed-out scan recovery retaining the last complete snapshot.

Test receipts are retained as dotnet-test-results (CI evidence, not release binaries).

Product version remains 0.1.3-beta.3. This change is merge-to-main only: no version bump, tag, release publication, container publication, or production deployment. Native Windows/Linux collection and deployed visual acceptance have not been run on the owner's installation; those remain checks for the future beta.