Well 10... e5 is not new or not a novelty at least. It is the cutting edge of theory right now and for all practical purposes white is crushing in these lines.
Unfortunately, I have a ridiculous load of correspondence games going at the moment so I can't get too deep into this line. I will try to return to it later but for the moment here it what a cursory overview of thoughts and findings.
10... e5
You didn't even mention 11. Bc3 which may just transpose to your lines but it bears its own investigation.
As for 11. Rd1 Bf5 12. Nf3 Ne7 13. Bc3 Nc5 14. Nxe5 0-0 this can be considered the mainline as all games with this 10... e5 move reach this tabiya. Although from here, white is scoring 70% so if you don't catch someone out of prep otb... you can be in a world of hurt defending this position imo.
Aside, what chessable course? Are you producing this course or consuming it? EDIT: it appears you are writing a course. Congrats! I would be happy to take a look at what you have. Of course, I need to now share a pet peeve which is to cite chesspub as a source. Because while you have given, you have also received. In doing so, you may bring more users here. I have in the past seen my work taken without citation for books and it caused me to contribute a lot less over the years.
B1 The only game I can find white won.
Regarding B2 Na4 certainly does not deserve a ! as it has lost quite a few games. Rae8 has held once. It seems the rook ending is probably lost after 23... Rae8 requiring 23... Re7 which is apparently enough to hold the draw though maybe improvements can be found.
In your follow up post regarding 9... f6 certainly 16... h5 is a vast improvement over 16... Kf7...
Jack Hughes wrote on 02/24/20 at 10:16:26:
I've also been taking a closer look at 18... Rae8. At one point it seemed like 20... Qg2 might be an improvement, but sadly it seems that white is quite a lot better after 21. Rf1 fxe5 22. Bxe5 Qxh2 23. Rd7! (23. Qxb6 Ng5 when black keeps the initiative and is apparently fine) when the almost forced sequence 23... Rf6 24. Re7 Rxe7 25. Qxe7 Rf8 26. c5 leads to an endgame where black is probably just objectively lost.
Somewhere I lost the thread on this one... what is the full line here?
Jack Hughes wrote on 03/12/20 at 01:24:51:
As regards to LC0 I should emphasise that I also have very poor hardware for running it, and so I am able to use it at anywhere near its full strength. Nevertheless the fact that it is just so different in style from Stockfish makes it very useful for analysis - it has helped me to discover several important lines that I definitely would not have found relying on Stockfish alone. On that subject, this includes several variations in the mainline Marshall Gambit, but sadly for white rather than black! I'm a complete idiot with computers so it took me longer than I would like to admit to start using it, but after watching a Youtube tutorial (link:
https://www.youtube.com/watch?v=J2srCBNK8jc&t=405s) I was able to work it out. Regardless of your hardware, I would highly recommend anyone interested in seriously analysing openings to do the same.
On that note, I do have very good hardware and have spent an awful lot of time in keeping up with hardware/software changes to play high level ICCF games. If you have a general question, feel free to PM me. I don't promise a response same day... but I do promise to respond with as much help as I can. When my game load decreases, I am also happy to offer some use of my high level hardware for lc0 for specific lines and ideas if it helps. Side note on lc0, it's pretty terrible at endgames and has a much shorter horizon effect. It finds very novel ideas... but they need to be checked thoroughly. And if I keep going off topic on lc0, I may need to start a new thread. Still one last thought, lc0 is primarily driven by which network you select (as the "brain" for lack of a better word). The best network is always in flux... they have a handy short list here that they keep updating for those without a bunch of time...:
https://github.com/LeelaChessZero/lc0/wiki/Best-Nets-for-Lc0A caveat though is that these tests are often done with specific purposes and settings in mind (such as submitting it to an engine tournament like TCEC). It is not done analysis partner or ICCF time control ideas... as such you really will need to do some of your own testing. Each network operates in some ways like its own engine. In the end, the easiest way to keep up on networks is to see which ones the devs are submitting for competitions.
Sorry I can't contribute more at the moment. Like I said, hopefully I can circle back.
As an aside, I don't share your optimism on the noteboom. When I worked as a second for the World Cup, I had it completely busted and my player went clear sailing to the next round easily on prep alone. I haven't checked it in some time... but I doubt all the holes on that opening can ever be fixed. I think it is super sharp and like praying your opponent isn't ready. But I would advise against recommending this to anyone as anything more than a surprise weapon.