Eastus2 shows a broad VM SKU regression across the basic and standard families, with many sizes flipping from available to unavailable in the monitor’s read-only catalog. The affected set includes basic a0–a4 and standard a0–a11, plus a long tail of standard DC-series entries. For SREs, this is a placement and fallback change, not a quota signal: capacity planning and SKU fallback lists should be rechecked.
The pattern is classified as a deprecation candidate, not a confirmed retirement, but the stability profile is harsh: 62 days of history, only 2 available days, 60 missing days, and 96.8% unavailable. Prior disappearances are zero, so this reads more like a newly observed delisting than a recurring blip. The last observed availability was 2026-07-12, with missing seen on 2026-07-10.
Coverage in eastus2 dropped from 2/64 regions previously to 1/64 now for many of these sizes, a net loss of one region in the monitored set. At the region-group level, North America coverage moved from 2 previous regions to 1 current region for the affected basic/standard sizes, so fallback choices are thinning locally.
Fallback placement still exists, but it is narrower. westus3 remains available for the basic and many standard entries, while eastus2euap also remains available for several DC-series sizes. That matters for latency-sensitive workloads and for regional evacuation plans where keeping SKU parity across paired or nearby regions is part of the design.
The underlying capability is VM size availability, which directly affects right-sizing, performance, cost, and capacity fallback choices. Reference: https://learn.microsoft.com/azure/virtual-machines/sizes/overview