For anyone who finds this thread and has the same question, sourceExtract is indeed the way to go here. What helps in this matter are the special regex characters used to look behind (?<=) and look ahead (?=). This enables you to capture exactly the portion of the code between two always-present elements of the web. The dynamic part of the web stored inside the variable by the sourceExtract can then be worked on by the usual Javascript in the executeScript_Sandbox command to get you exactly what you want.
Here’s a thread that gave me the right pointer: