When considering law, it's always worth noting that the specific particulars are arbitrary and path-dependent. I think it'll be hard to draw any kinds of conclusions on this ruling (which doesn't find against ABP, merely kicks the issue down to the lower court for reconsideration, not unlike Oracle v. Google with regards to API copyrightability) without reading the whole thing.

(One piece in particular I'm personally naive on is what German legal precedent says about consumer's right to modify consumed material. In the US, an author's copyright doesn't stop me, the reader of a copy I bought, from highlighting the book up, or crossing out passages I don't like, or tearing pages out, or turning the thing into a delightful booksafe. Naively, I'd believe ABP should be considered in that category of thing: an accessibility tool people use to modify the material they consume to better fit their needs. It doesn't modify the author's original work and it doesn't grant the reader the right to transmit the modified work to someone else, so I'm unclear on how copyright protection should be thought to enter in here, and I bet the text of the ruling clarifies).

Yes, the particularly odd thing to my eye is the lack of redistribution. My locally modified page infringes even though I am the only one who has seen it. Odd too that many kinds of blockers don't actually rewrite the markup at all, they just don't fetch assets from blocked endpoints, which doesn't seem like 'modifying the code' somehow.

this is my confusion also - it sounds like they are conflating "modify and sell" with "modify and consume"