Cross attention - is our prompt scheduler really giving same results ?

[cmp-nct]

Here is (one of the many) feature requests that was closed as 'implemented':
#422

The only implementation I could find that somehow goes into that direction is the prompt scheduler.
So you can make a prompt like this:

A big plate with spaghetti with [sugo:30] and parmesan [and some basil leafs:30]
A big plate with spaghetti with [sugo:1] and parmesan [and some basil leafs:1]

This works somewhat but not anywhere close to the many examples I have seen on the Cross Attention Control implementations around. The above example adds more "fork" than basil to the spaghetti ;)

The second option is the noise finder, it works well for some things but fails on others.

So are we really at a comparable state than the others in that regard ?
Did I miss something ?