Oh dear! Thank you for the bug report. I'm investigating right now.
The issue is now fixed and V2 of the app is uploaded. The problem was triggered by a short stack of exactly two coins falling in unison. The first would hit the tail off a switch and fall through and the upper coin would then get caught (but the slot would remain green). An interesting edge-case, although I'm surprised it crops up so rarely!
Thank you for the quick fix.
I am in the process of making a strategy. I find it hard to reason about the timing of which coin falls down first when multiple fallings are happening. (Maybe I'll take a peek at the source code 🤔)
The strategy beyond two levels deep I really struggle with also. If you can find a solid strategy for the 7-switch puzzle I would consider this a success, especially if it also works for the 10-switch puzzle. Historically this puzzle was first offered as a java applet (20 years ago) which in itself was a virtual manifestation of a pre-existing physical puzzle, neither of which I originally implemented. The bug yesterday made me question whether this new app is behaving exactly like the old applet. Both apps must be fully deterministic, but differences in timing and animation could mean events happen in a slightly different order, during a large cascade of coins. The available coin slots at the top would have been chosen to ensure that all puzzles are solvable regardless of the start-state of the switches, but is this still true? Now I am not so sure. If I can clarify, I will let you know.