Skip to content

AZ-140 deep review — September 28, 2026

The AZ-140 guide was read in full and mapped to 78 objectives in twelve groups. It now has seven worked examples, ten labs and 48 explained original checks. Thirty offline assertions verify the calculations. No Azure, tenant, Arc, endpoint, media, storage, update, backup/restore, failover or paid-content execution occurred. Independent human review remains pending.

Scope and lifecycle

The official blueprint retains its July 20, 2026 baseline. Objective/status hashes are unchanged and no future objective announcement is shown. The credential is active, with 100 minutes, seven languages and renewal. Practice Assessment and sandbox links were verified at the landing page only.

The review separates four timelines:

Item Current primary evidence Applied treatment
AVD classic September 30, 2026 retirement blocks classic connections Distinguish service generation from the active certification
Windows Remote Desktop clients Public-cloud MSI support ended March 27; government/21Vianet/classic extension ends September 28; Store app ended in 2025 Inventory executable/distribution and move to supported Windows App
Classic Teams VDI availability ended July 1, 2025 Do not teach classic Teams as a supported fallback
Windows-endpoint WebRTC optimization Support ends October 1, 2026; availability ends April 1, 2027, for the announced AVD/Windows 365/Citrix scope Separate app installation from verified SlimCore and call/sharing evidence

The AVD Teams overview conflates the classic-client and optimization deadlines and differs from the Teams VDI requirements on some client minima. The guide explicitly uses the dedicated timelines and requires a currently supported client distribution; it does not invent a reconciled minimum version or claim Microsoft corrected its page.

Applied technical changes

  • Management: distinguish the creation-time host-pool management choice and service-managed configuration from external standard-pool tooling. Host replacement needs surviving capacity, externalized state, monitoring and tested rollback.
  • Identity: update the universal service-principal assumption for managed identities, retaining App Attach's exception and feature-specific authorization.
  • Architecture: add Hybrid boundaries and its unsupported multi-session/power-management features. Clarify Multipath's client/cloud scope and separate private discovery, feed, host connection and storage endpoints.
  • Profiles: apply current Entra Kerberos requirements and the profile procedure, with scoped storage-app authentication exceptions. Record FSLogix 26.08 and its unchanged ticket lifetime, plus provider/cache recovery limits.
  • Applications and scale: distinguish package/attach/registration/launch and service-principal storage access; separate dynamic scaling from power management, exclusions and personal-host hibernation constraints.
  • Servicing: use the OS/update-type matrix rather than carrying one monthly-update recommendation into every OS upgrade. All seven tables were inspected with cell boundaries preserved.

The worked examples cover maintenance capacity, network demand, sign-in I/O, per-file handles, cache flush time, migration-evidence denominators and full recovery timing. The numbers are original assumptions, not performance claims, actual autoscale implementation or proof of recoverability.

Learning resources and blog intake

The Microsoft course lists four days and four languages; its loading syllabus did not reproduce the old self-paced runtime. Only the first readiness episode was inspected, with chapters through 18:16; the old series total is qualified. The added Pluralsight path offers six Ned Bellavance courses dated 2023–2024. Their durations total 9 hr 14 min against a rounded nine-hour header; it is a structured foundation requiring current-doc reconciliation.

Direct O'Reilly/Udemy access remains blocked. Indexed primary listings confirm Mastering AVD (July 2024, 718 pages, 14 hr 54 min platform estimate), Securing Cloud PCs and AVD (June 2024, 396 pages, 8 hr 22 min), and Kubaib's course (April 2026, 11 sections, 125 lectures, 23 hr 24 min). MeasureUp lists 130 product questions and a December 2024 update; its generic FAQ says about 150. Product-specific metadata replaces the guide's generic count. No paid material, actual questions or completeness/originality audit was performed.

Microsoft blog Reading boundary Original learner task
Enhanced host management, NeoCai, July 9 Main announcement and short setup outline; linked code not inspected Management ownership and maintenance-capacity worksheet
Hybrid GA, Steve_Downs, September 1 Architecture, ownership/licensing and deployment outline; promotional claims excluded Supported-feature comparison and responsibility diagram
Teams VDI transition, Fernando_Klurfan, March 2 Timeline, scope and quickstart checklist; later pitfalls only partly read Endpoint-scoped deadline and adoption-evidence worksheet

Public embedded article bodies supplied content where rendered pages were shells. The receipt records author/date metadata, extraction hashes and read boundaries. Announcements were checked against implementation guidance; marketing benefits were not adopted as measured results.

Validation and follow-up

The receipt at ADLC_Docs/operations/2026-09-28-az-140-deep-review.json contains objective mapping, source results, historical evidence, guide hashes, bounded reading notes and actual gate results. Thirty assertions cover all numerical examples, private endpoint counts and learning-time arithmetic. Unchanged objective/status snapshots are preserved, alongside the earlier review record.

Repository tests, catalog consistency, repository validation, strict site build, generated-site validation and whitespace checks gate publication. Their completed outcomes are recorded in the receipt. Events schedule checks for classic/client retirement on September 30, WebRTC support on October 1, management/profile currency on October 12 and the later optimization availability deadline on March 1, 2027. Automation detects and queues changes; semantic review, lab execution and human judgment remain separate work.