DECISIONS: Provide guidelines on how all Shield add-on studies should behave in general cases

[motin]

Moved here from mozilla/shield-studies-addon-template#68

From Krupa:

would it be possible to have general guidelines on how all SHIELD add-on studies behave when it comes to uninstall/remove actions by users?
i see a long email thread about what the expected behavior should be for jonathanKingston/http-dns#13

My thoughts:
The v4 Shield Utils API takes care of uninstalling the add-on if the user opts out of the study via an API method; however the study Krupa linked does not use the Shield Utils API, which explains the discrepancy. I agree that it would be:

• good to have this documented somewhere, and possibly
• require that all devs use the Shield Utils API for Shield studies.

Either way, we have been discussing whether or not to uninstall the add-on when the study ends, and those kinds of decisions and rationales should be documented somewhere. Establishing these guidelines may require some collaboration with UX, data, legal, etc.

[motin]

Agree, we should consolidate a list of shield-relevant behavior and decide what is the appropriate behavior. Some issues asking about this are the following:

• Should we leave private browsing users alone? Leave users alone if they are in Private Browsing mode shield-studies-addon-template#51
• Uninstall on disable? Is this the expected behavior? The shield study is not removed when the "Disable" button is clicked #231, Despite claims to the contrary by me, Disable doesn't actually force uninstall. #247
• Undo-option after study add-on is removed? Is this the expected behavior? The "Undo" option is wrongly displayed after the shield study is removed #230
  There is also a proposed engineering plan to solve the uninstall/disable/undo-related behavior: "Better Study Endings" Engineering Plan (Proposed) #246