Playing with your experimental line numbers pde.jar (I took a quick look at it in a HEX viewer, but didn't notice anything obviously a revision number in the first FF bytes. This one (for future reference) is 833,721 bytes long. Shall we call it line_number_beta01?)
Horizontal scrolling is broken. (Oops...)
As I mentioned before I like the changing of line numbers for lines with comments. But with for one of my actual sketches where I have some lines that are comments only and some lines that have a comment after a bit of code, the line numbers are all comment color. I'm kind of torn here. The current way helps highlight where the comments are, but when scanning code to try to find the line number where the error code is it is easier to skip over the greyed out line numbers. I'd like to suggest that the line number only changes to grey (or whatever the comment syntax highlighting color is) when the only contents of the line is a comment. But what about scanning the code just to find comments. Which is more important? Not sure how to decide this.
Oh... I just figured out how you did the coloration of the line numbers. It matches the syntax highlighting of the last character of the line. Clever. I think it might make more sense to use the syntax colorization of the first non-white-space character of the line... Dunno. That would be a judgement call based on playing with both techniques. (Though the current method, except for comments, might help highlight lines where one forgot to end it with a semi-colon. It wouldn't catch all of them, but might catch some...)
I think I do prefer left-justified line numbers, but that is just a very minor concern. I wouldn't think less of this effort because of the line number justification. I just think it is easier to see way down in the code which is the top level closing bracket when closing nested blocks if there is always 1 space between the line numbers and the code.
It's tough to tell (I hate counting multiple white-space characters, can't tell one from the other...), but it looks like you have 4 characters allocated for line numbers. When I went over 5-6000 lines I saw the line number area increase by 1 character to make room for 5 digit line numbers (at that point I was just pasting blocks of 1000 lines of garbage characters to test the line numbering). A bit of fine tuning and I found the jump happened when I pressed return to create new line number 5013. Not a criticism, just a comment on my experimentation so you can verify that it does or doesn't work as planned.

Starting with 5001 lines and inserting blocks of 1000 lines at the beginning of the file didn't get the extra digit added to the line number area until the third paste (expected it on the first paste of 1000 lines). Still not a problem, but might point to a small instability somewhere.
Except for the horizontal scrolling not working, I haven't run into any actual bugs yet. I checked the "option to not move the carat" pde.jar and the horizontal scrolling works there. I hope this helps.