Whe build pods fail due to running out of memory, the pod status is reported "Error" instead of "OOMKilled" (or similar)
    

4.15
    

Always for builds in 4.15, not currently observed in 4.16 or earlier versions (4.14)
    

    1. Launch OpenShift cluster v4.15.z with Build capability enabled (the default)
    2. Create project/namespace
    3. Create BuildConfig that runs build with low memory request/limit. See https://raw.githubusercontent.com/openshift/origin/master/test/extended/testdata/builds/statusfail-oomkilled.yaml for an example
    4. Start the build (with example above, run `oc start-build statusfail-oomkilled
    

    Build fails with generic error reason, but pod exit status 137
    

    Build fails with OOMKilled reason 
    

    Observed in gcp-ovn-builds test suite for 4.15.
    https://github.com/openshift/origin/pull/28725 and its backport will temporarily unblock us - these PRs should be reverted once a root cause is identified and a fix is delivered.