Likely due to step_name logic in the swarming recipe module:
https://codesearch.chromium.org/chromium/build/scripts/slave/recipe_modules/swarming/api.py?rcl=a9113e8e59271b3d95d1354389dec5bab3725fbb&l=1223

I know the GPU steps have its own step_name logic as well:
https://codesearch.chromium.org/chromium/build/scripts/slave/recipe_modules/chromium_tests/steps.py?rcl=a9113e8e59271b3d95d1354389dec5bab3725fbb&l=1325

Problem likely lays in how those two interact.

And the ios steps use their own recipes, so their problem is probably buried in there somewhere.

I can try to resolve the difference. Do we want to keep "step_ui_name" the same and change "test_results_step_name"? Or do we want to keep "test_results_step_name" the same and change "step_ui_name"?

Probably the former since the values for "step_ui_name" are more descriptive.

It seems that we have some inconsistency in step names in a few places. It would be great if we could unify them.

For this case, I would suggest keeping "step_ui_name" as it is, and change "test_results_step_name" to be the same as "step_ui_name".

A better solution for this might be to:
* add a numeric id to each step in the build and surface the ids in all BQ tables for data joining
* also add a metadata to each step to make it possible to group the same test of different steps (with patch), (without patch), & (retry summary).

This needs a bigger change in all the places though.