This only happened once in a dev submit run with 19-b17 on OCI aarch64. I have not been able to repro it in over 25 tries.

Errors in benchmark: scimark.lu.small
  [warmup] Harness interruped during measurement period.
  [warmup][bt:23|op:234] Validation failure on line 1.
    Expected output:
    [001] The result is correct.
    Received output:
    [001] The result is wrong. The result is 1.2270295890459693E-11 instead of 1.3607136251092555E-11
    [002] The result is correct.