Been a long time since I last worked on LoopDeLoop, but felt like practicing some AI coding outside of work so I've refreshed a bunch of my old stuff.

https://themissingdocs.net/loopdeloop/ is the new website version of loop-de-loop - I don't pretend that the UX is great and it doesn't have every feature of the old winforms version.

It does however have 10 more layout options than the old version - based on the slitherlink variations types on this web page - in case you want to practice more problems than are available in the archive...

I'm loving it so far, especially the smoothness of applying the lines. I did notice a couple of bugs playing it last night.

1) In a few rare instances it wouldn't let me apply a line thinking it was invalid when it wasn't. When I did an undo and tried it again it worked.

2) For one puzzle that took me a while, I had solved it but it didn't pick it up as a solve even though every line was valid. I took a screenshot and then hit "solve" and it matched my screenshot.

for 1) I'm assuming you are using the disallow bad moves option.  I'm not a big fan of it myself so haven't tested it much. I recall long ago having some similar issues where the internal state gets corrupted because a disallowed action isn't correctly rolled back.

(Just to be clear it disallows moves which would cause the puzzle to immediately be in a state with a clear violation of the rules, it doesn't disallow moves which are not a match for the final solution... I'm assuming you aren't being confused by that.)

2) I'm guessing is related, if the internal state gets corrupted in the right way it'll be unable to recognize that the loop was solved correctly.

I asked my AI session to do a quick investigation and it thinks the issue could possibly have been to do with puzzle state being preserved over browser refresh incorrectly, or with having previously used 'solve' and then undo on that solve. I've pushed its fixes (which do appear valid), if you continue to see issues I can try and check again. (Although I'm low on AI quota at this point )