Court of law
If you were relying on a document in a court of law and it had been signed with SHA1, then my assumption is that if two parties produced two documents with the same hash, there would be some examination of the documents. The guess is that the original would be 'clean' and the doctored one would have the doctored text (the subtle addition of the word 'not' where appropriate) PLUS some additional garbage to bring the signatures back into line. As it's astronomically unlikely to produce the same signature when only making a meaningful semantic change to the text, it should be possible for a forensic examination of the documents to identify which is 'clean'.
However, if SHA1 is only being used in an automated system (Git etc.), then the issue is much more pressing - the assumption being that the system will not be coded to identify a collision, which would be processed and actioned unnoticed.
Biting the hand that feeds IT
