An Azure service that provides an event-driven serverless compute platform.
Thanks Vincentius.
@Gursimran Singh — appending more precise evidence in case it helps a platform-side investigation, and also closing out this thread since we've resolved it on our end.
More specific correlate than the host version alone: within host 4.1052.300.26370, the failure was isolated to instances that came up running Node 18, even though our Function Apps explicitly declare node|20 in functionAppConfig.runtime. In the same window, 942 Node-20 worker starts on the identical deployed package had zero errors, while 187 Node-18 worker starts all failed identically at load with crypto is not defined (thrown from randomUUID() at module-import time in @azure/data-tables → @typespec/ts-http-runtime — Node 18 doesn't expose globalThis.crypto, Node 20+ does). So the rollout appears to have started some workers on a Node major the app never requested, and any app depending on globalThis.crypto at import time would hit this the same way.
For anyone else who lands here: we didn't wait on a host-side fix. We migrated our Function Apps from Node 20 → Node 22 (Node 20 was already EOL as of 2026-04-30), which sidesteps the mismatch regardless of which host version an instance lands on. Recovered ~3 hours after onset. Confirmed, as you and others noted, that Flex Consumption doesn't support FUNCTIONS_EXTENSION_VERSION or linuxFxVersion pinning — upgrading past the runtime version at risk was the only real lever, not a platform-side wait.