We starting seeing failures are all jobs that use the nightly image starting on March 5th due to changes in the use of sigstore. The nodes install correctly and the images are written properly but on reboot we see the logs flooded with these sigstore related failures:

Mar 10 17:51:22 master-0 podman[2457]: 2026-03-10 17:51:22.055968197 +0000 UTC m=+0.200357680 image pull-error  quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:15ec9f88180ce7d7e9ff53418a13a7d6a36f397eb4e297257cdc89613c574325 unable to copy from source docker://quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:15ec9f88180ce7d7e9ff53418a13a7d6a36f397eb4e297257cdc89613c574325: Source image rejected: A signature was required, but no signature exists

This is occurring with all agent jobs SNO, Compact, TNA etc. that use nightly images e.g.

registry.ci.openshift.org/ocp/release:4.22.0-0.nightly-2026-03-09-103941

When using the ci images it works fine e.g.

registry.ci.openshift.org/ocp/release:4.22.0-0.ci-2026-03-10-023220

It looks like we'll need to set this env variable for assisted, although its not clear if there is currently support in assisted for it:

OPENSHIFT_INSTALL_EXPERIMENTAL_DISABLE_IMAGE_POLICY=true

We did see that this field is set for the tests here:
https://github.com/openshift/release/blob/882d7cf0cd312ba2cda2089a8af9d18f8da71e05/ci-operator/step-registry/baremetalds/devscripts/setup/baremetalds-devscripts-setup-ref.yaml#L26-L28

There's more info related to the agent-installer in this thread:
https://redhat-internal.slack.com/archives/C02SPBZ4GPR/p1773099342126439

For reference, this was a recent fix in the core installer that worked around this issue:
https://github.com/openshift/installer/pull/9940