Ad 7) To my opinion there is some kind of bug concerning "word recognition". My opinion is that the algorithm treats special letters with accents (but not all, only some) as "not a letter" which means that it cuts the words to smaler parts (it happens a lot in Czech texts).
In your case, the reason for ill behaviour is that spaces are missing. There is a ill-behaviour of annotation export for certain pdfs. If I am not mistaken, LaTex generated. All spaces in text disapear...
I tested dictionary a bit.
So, the (abc) bug is solved for most of dictionaries in 1.8. But not all. I have one dictionary which uses a lot of font formating and it is showed without removing or correct interpreting.
The bug "not recognized dictionary change" is still there. Strange is that change of dictionary is immediately reflected in (abc) view, but not others.