In puzzle 2059, create one or more bands.
Enter ligand design mode. Make a change to the ligand, such as adding or deleting an atom.
The bands disappear.
Getting the bands back requires three undo commands (control-z or Undo button). The undo graph seems to show two extra steps.
The multiple undos with the ligand design puzzles is a known issue. There should be a fix to it coming out with the next devprev release.
Bands disappearing when you modify the ligand is also something which is not unexpected. Keeping track of which atom is which when the molecule changes (to figure out where to put the bands) is somewhat tricky. We haven't quite figured out how to robustly match things up in the general case, so just clearing the bands to the ligand is the (temporary) "safe" option.
Perhaps you can approach the disappearing bands problem as below:
Have a Save Ligand Bands button. It takes the starting ligand structure with its bands and treats it like a guide. It then converts the bands to the ligand into bands to space, perhaps giving them all a special color. Then the user can make changes to the ligand structure. When the user is ready to redo the bands to the ligand, they press a Restore Ligand Bands button. This button looks at the positions in space that each band points to. It compares atoms very near these endpoints with the banded ligand atoms in the original structure. If there is enough of a match, it changes the band to space back into a band to the best-matching ligand atom. If there isn't enough of a match, it could leave the band to space alone, reminding the user to either remove the band or connect it to an appropriate atom by hand. Perhaps it could color the dangling bands to space a special way to make them easier to find. Perhaps the Restore Ligand Bands button could have two settings: (1) remove bands that don't give good matches to the original structure or (2) leave alone the bands to space that don't give good matches to the original structure. Option (2) then reminds the user to delete or reconnect appropriately the remaining bands to space.
I'd imagine a similar approach could be used to preserve bands to sidechain atoms when one mutates a sidechain. Perhaps for sidechain atoms, the Save/Restore Sidechain Bands buttons only operate on selected segments instead of on all of the protein's sidechains. I'd imagine many bands to backbone atoms would be preserved this way. Perhaps bands to certain sidechain atoms would be preserved as well (like when one mutates ASN to ASP or GLN to GLU).