I don't think this is actually a bug. The issue is that the config settings such as doApplyExternalPhotoCalib=True must also be specified for the subtasks:
config.assembleMeanCoadd.doApplyExternalPhotoCalib=True and config.assembleMeanClipCoadd.doApplyExternalPhotoCalib=True . An alternative would be to check the configs of the subtasks when they are created and copy over all of the settings from the main Task, though I would expect that to create its own problems.

The original code ensured that the the coaddMean and coaddClip were generated with the same configs as specified for the final coadd, except for MEAN/MEANCLIP.

        configIntersection = {k: getattr(self.config, k)	

                              for k, v in self.config.toDict().items()	

                              if (k in config.keys() and k != "connections")}	

        config.update(**configIntersection)

        config.statistic = 'MEAN'	

        task = AssembleCoaddTask(config=config)	

        coaddMean = task.run(skyInfo, tempExpRefList, imageScalerList, weightList).coaddExposure	

        config.statistic = 'MEANCLIP'	

        task = AssembleCoaddTask(config=config)	

        coaddClip = task.run(skyInfo, tempExpRefList, imageScalerList, weightList).coaddExposure

When you replaced these with subtasks, you forgot to load the camera-specific configs for the two subtasks.
This changed the behavior of safeclip for all cameras that have an obs_camera/config/safeClipAssembleCoadd.py
We didn't notice until now because we don't regularly run safeClip in CI.

Please either revert your ticket or add the following:

config.assembleMeanCoadd.load(os.path.join(os.path.dirname(__file__), "assembleCoadd.py"))

config.assembleMeanCoadd.statistic = 'MEAN'

config.assembleMeanClipCoadd.load(os.path.join(os.path.dirname(__file__), "assembleCoadd.py"))

config.assembleMeanClipCoadd.statistic = 'MEANCLIP'

to every obs_package that has a config/safeClipAssembleCoadd.py

I'm happy to do this ticket myself, but I would have just reverted https://github.com/lsst/pipe_tasks/commit/09a6bcfc91ceffd0e07f70b230a29e7cbca81b13#diff-d2c8262f4554a7b63f3ea59d253e64f595e029ac06f1b031eca6cf0ab6d486fd

I couldn't tell what problem that commit was trying to solve, so I wanted you to take a look and see if there was a solution that solved your problem AND maintained original config behavior.

That change was required in order to create Mock versions of the various AssembleCoadd Tasks for unit tests. I find it valuable to have those tests, so I will follow your suggestion to add the camera-specific configs to the obs packages.

Note: it looks like only obs_lsst and obs_subaru have camera-specific configs set for SafeClipAssembleCoaddTask.

I modified the two obs packages that provide configs for SafeClipAssebleCoaddTask, but I also tested reverting the changes that caused the problem in the first place. The unit tests that I was concerned about pass even with the changes reverted, so I am happy with that solution. For completeness, I made pull requests both for the obs package changes and for reverting pipe_tasks, but the two options are not compatible. Both pass on Jenkins.

Obs package changes:
https://github.com/lsst/obs_subaru/pull/332
https://github.com/lsst/obs_lsst/pull/278

Reverting pipe_tasks:
https://github.com/lsst/pipe_tasks/pull/446

Please choose your preferred solution as part of the review.

I lost track of this ticket, and am coming back to it now. I believe the correct solution is to just revert the changes to pipe_tasks, so I have rebased that branch and am checking that it passes on Jenkins. Yusra AlSayyad could you verify this is still the desired change?