in #3805, the behaviour of feaLib was changed so that mixed single-and-ligature rules would be combined into a single ligature lookup.

This can interact with the behaviour of the aalt feature. Consider the following example:

feature aalt {
    feature clig;
    sub x by z;
} aalt;

feature clig {
    sub a by b; # we expect this to be in aalt, but because it gets compiled
                # as a ligature rule this doesn't happen
    sub e l f by z;
} clig;

Here, per the spec, we expect that the single sub rule in the calt feature is added to aalt; however since we are actually compiling it as a ligature substitution, this doesn't happen.

Do we care?