Instead, recommend people to use a save file instead.
The limit could likely be tested by creating as many fake entries as possible. A bisect method would be useful, starting with an extremely high amount of reflections to find where it doesn't work. Then decreasing to an amount where it does work, and go up again from there.
Edit: this will also need to apply for the QR code. For larger datasets, only allow file export and import.
Instead, recommend people to use a save file instead.
The limit could likely be tested by creating as many fake entries as possible. A bisect method would be useful, starting with an extremely high amount of reflections to find where it doesn't work. Then decreasing to an amount where it does work, and go up again from there.
Edit: this will also need to apply for the QR code. For larger datasets, only allow file export and import.