Duplicate instance of related action.

These are not duplicates and is working as intended since every time you click on a related action the desired behavior is to initiate an action. I don't think you can use the fact that the relation was launched from the same browser to avoid initiating a new process. For defining duplicates you need to have a piece of info e.g. the record ID and if you know it should have no more than one running instance of that related action then you can use a DB flag to hide the related action for a specific record once it has one started. In addition you can also add logic to the beginning of your process to check if there is already an existing process instance for that record ID, if so just terminate the new process without doing anything else follow the normal flow. This should catch the edge cases where the related action is triggered in quick succession or if someone directly paste the URL like you mentioned.

These are not duplicates and is working as intended since every time you click on a related action the desired behavior is to initiate an action. I don't think you can use the fact that the relation was launched from the same browser to avoid initiating a new process. For defining duplicates you need to have a piece of info e.g. the record ID and if you know it should have no more than one running instance of that related action then you can use a DB flag to hide the related action for a specific record once it has one started. In addition you can also add logic to the beginning of your process to check if there is already an existing process instance for that record ID, if so just terminate the new process without doing anything else follow the normal flow. This should catch the edge cases where the related action is triggered in quick succession or if someone directly paste the URL like you mentioned.