Remember also that the fragment isn't passed to the server, so there's two things that need to happen for them to work:
1) The fragment-free URL needs to contain enough information for the server to load the correct page (for questions with multiple pages of answers)
2) The fragment itself needs to contain enough information for the script to identify the correct *post* (since looking the comment up by ID won't work if it isn't among the comments loaded by default), load the full set of comments if the linked comment isn't visible, and then find the comment (so both post ID *and* comment ID).