I remember that one of my most recent updates to my site was on Halloween and featured an article on the Halloween Gambit. I still need to update the site to address the fact that some pages are in disarray, but so far "life" has been getting in the way of me completing that. I hope to manage it at some point in the next few weeks.
With it being Halloween, I've completed my first complete Lichess study (I've also got a couple ongoing on the Scotch Gambit and the Blackmar-Diemer Gambit) which is on the Halloween Gambit.
My conclusion is unchanged though from last time: it's dangerous, but unsound. I recently played in a Chess.com thematic tournament where I played the 7...d5 line with Black and converted the extra piece, although my opponent missed an opportunity to complicate matters. When I played White, my opponent went 7...Bb4 instead, and I had an opportunity to generate a strong, if not winning, attack, but I missed my chance and lost.
As I mentioned in the Introduction, Black can try to transpose to the Stafford Gambit (1.e4 e5 2.Nf3 Nf6 3.Nxe5 Nc6) with 4...Bc5. Objectively the transposition with 5.Nxc6 dxc6 is good for White, but I'm inclined to suggest 5.Nf3 instead, which allows White to retain an extra pawn without permitting Black as many hacking chances as in the Stafford Gambit. I'm currently playing in a Chess.com thematic tournament in the Stafford, as it happens (the first time that I've ever played it with either colour), and have so far got good positions with both colours, but it's still early days there.
A good source on the Halloween Gambit and the history of it is Tim Krabbe's site at https://timkr.home.xs4all.nl/tour/breeze.htm My investigations seem to bear out the old masters' view that the lines ending 7...d5 and 7...c6 (Chapters 9 & 11) are the biggest test for White, but White still gets some (but objectively not enough) long term attacking chances there.