Peter Bjuhr wrote
Based on these two examples, it would indeed seem that LilyPond is
intelligent enough to adjust for this potential problem; in the first
example where there were no risk of mistakenly take the accidental for a
part of the key signature the distance were 1.3 and here when the risk
is apparent it is 1.5. If this is the case and not just random
coincidence I would completely drop this from the list of potential
problems. What do you think Gilberto?
In the first case, there is no risk of mistaken the flat accidental on the
aes for a key signature, but on the bottom case it has the risk. And both
have pretty much the same spacing (around 1.4 units). So I don't think
LilyPond is adjusting for potential misunderstandings.