I still haven’t seen a convincing case for why you would want duplicated custom fields on sub-tasks in the first place, but here’s an example of why this is causing issues:
One of our projects is used to manage our content queue. The main task is the actual article, and the custom fields are things like word count, primary keyword, and current stage (there are about 10-15 of these). Sub-tasks are things like keyword research, write first draft, upload article etc.
With this latest change, each sub-task now has a long list of empty custom fields that are irrelevant to the task that staff need to scroll through just to get to the actual task information, which is contained in their own set of sub-tasks. Information is also now getting filled in at the sub-task level rather than the primary task by mistake, which confuses things further.