Using pipelines and golden/shared datasets (Live query)

Hi All, 

I'm using the golden dataset model of Live Querying datasets in a 'dataset workspace' in reports sitting in a separate 'app workspace' (See https://www.jamesserra.com/archive/2020/06/power-bi-faq/ first paragraph (hub/spoke). 

I'm also trying to use pipelines for the dataset and the reports. So our developers can make changes to the reports and datasets separately, other developers can review and approve and each can be deployed independantly. 

The issue I'm coming up against is shown in the attached image. In short, when publishing a thin report to production, I can't  set it to reference the production version of the 'same' dataset. That doesn't seem to be a feature and perhaps I should submit it as an idea. 

You might say..'why don't you just set the report to the production dataset in powerbi desktop and upload it that way to the test report workspace and deploy (or direct to the production report workspace instead of deploying). 1. I'd rather the test report always point to the test dataset and 2. It doesn't work because when you change the live query to a different dataset, it changes the report enough that the published PBIX (despite having the same name) is published as a new report and will not overwrite the old...losing bookmarks, cached filters...all that stuff. So..yes you can change the Live Query but unfortunately it breaks a bunch of other stuff. 

This is solved to some degree by using an API (https://docs.microsoft.com/en-us/rest/api/power-bi/reports/rebind-report-in-group) call to re-bind a report to a different dataset in the service directly. Here is a person describing that problem. 

It is surprising to me though that Golden datasets seem to be the recommended method of using Power BI and must be used by many people who would also like to use pipelines and there seems to be this gaping hole in functionality. Am I missing something? Does anyone solve this differently? 

I'd rather not have to write up work instructions that include API calls which involve copy pasting elements of URLs and opening up Powershell to complete what should just be a click of 'deploy to production'. 

I know we can switch out data sources and parameters for datasets when deploying between test and prod. Why not reports which have live queries as data sources? Thanks

Pipelines and golden datasets