I promised to do this already a few days ago, but couldn't find time for it. Sorry!
So here is a list of all line-initial words in the VMS. It's taken from the Takahashi transcription (I haven't checked the accuracy of anything, so there are probably some errors), excluding single-word "lines", most of which are actually labels.
From left to right, the columns are:
- the word type
- number of line-initial occurrences
- number of all occurrences
- proportion of occurrences which are line-initial

There are in total 1958 word types, including:
- 1000 unique ones
- 114 others that occur only line-initially
- 844 that are not exclusively line-initial

I hope this will be helpful.

I've done this already a week ago and much much more  (and published it. Since there are no people interested and I also  I've made some conclusions, I decided to take it offline)

Strange numbers you have. I've used only lines not labels, because you can not call such a text "a line"

I have:

3905 filled lines
of which from the unique list of words overall 1771 words
of which with only 1 repeat: 1299

The words that are position-unique were discussed already.

Davidsch: I have 4354 lines. Maybe the difference is because there are some labels consisting of several words, which my script would have counted.

I checked the top of the file and it is identical to my results, with respect to the previous remark.

This point is indeed interesting and also it seems that the first word of each page seems to be of less frequent than average