VS402984: Pipeline fails when referencing missing agent pool

Goal

Determine the most cost‑effective agent strategy for Azure DevOps pipelines that run only a few times per month.

Constraints and uncertainty

The VS402984 error (“The requested resource does not exist”) appears when a pipeline YAML references an agent pool that has been deleted or never created, underscoring the need for a valid pool even for low‑traffic workloads. Switching from Microsoft‑hosted parallel jobs to a self‑hosted agent on an Azure B‑series VM can lower per‑minute charges, but the VM incurs uptime costs and requires patching, hardening, and monitoring. Enabling pipeline caching reduces consumed minutes by reusing dependencies, yet the storage overhead and warm‑start latency for infrequent runs are not quantified in the documentation.

Open questions

Does enabling pipeline caching provide a net cost saving for pipelines that execute fewer than five times per month?

What is the storage cost impact of retaining cache artifacts for long periods in a low‑frequency scenario?

At what monthly pipeline‑minute threshold does a self‑hosted B‑series VM become cheaper than a Microsoft‑hosted parallel job?