[actions] possible regression, jobs are not complete but canceled instead

[nobuto-m]

I started to see a change in the testflinger action behavior recently.

For example,

  + exit 1
  2025-05-09 11:31:04,608 hinopio WARNING: DEVICE CONNECTOR: Tests failed, rc=1
  2025-05-09 11:31:04,609 hinopio INFO: DEVICE CONNECTOR: END testrun
  
  *************************************************
  
  * Starting testflinger cleanup phase on hinopio *
  
  *************************************************
  Cleaning up container if it exists...
  hinopio
  2025-05-09 11:31:11,557 hinopio INFO: DEVICE CONNECTOR: MAAS: Reading node's block device information
  2025-05-09 11:31:13,521 hinopio INFO: DEVICE CONNECTOR: MAAS: Releasing node hinopio

^^^ a test has failed at this point and MAAS machine is going to be released.

}2025-05-09 11:31:49,541 hinopio INFO: DEVICE CONNECTOR: MAAS: Successfully released node hinopio
  Error: The operation was canceled.
Run echo "::group::Cancel job"
  echo "::group::Cancel job"
  testflinger --server $SERVER cancel $JOB_ID
  echo "::endgroup::"
  shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0}
  env:
    SERVER: https://testflinger.canonical.com/
    JOB_ID: d5d5133f-d4cf-4993-b1cf-4ed78b013f91
Cancel job

^^^ successfully released machine, then Error: The operation was canceled.

[syncronize-issues-to-jira]

Thank you for reporting your feedback to us!

The internal ticket has been created: https://warthogs.atlassian.net/browse/CERTTF-627.

This message was autogenerated

[nobuto-m]

The only change I can see is: #605 but not sure if it's the root cause or not.