I do not think it is a problem since the concept of an approaching collapse is still present and a driving mechanism in what we are trying to describe. When a multiplex is about to take place, that specific approaching collapse reached collapse, but in the pattern as a whole, there is still other approaching collapses active to keep the chain reaction of events going. I do not think it is a problem, but it is an important difference to note.
There is the macro approaching collapse in the situation of an incoming object towards a hand already occupied by a object, but there is a zoomed out approaching collapse that compromises of all those macro situations as a whole. The pattern as a whole so to speak. So we can talk about the approaching collapse of a specific object, or the approaching collapse of the pattern as a whole. The first is defined by an object reaching a hand already occupied by an object. The second is defined by a constant presence of at least one approaching collapse somewhere in the pattern. As long as there is at least one approaching collapse still active, the link of new approaching collapses triggered by saves, remains unbroken.
There are three different concepts that we need to differentiate between:
“Vanilla” juggling - all links remain unbroken (ex. 333, 531 etc)
Multiplex juggling - at least one link remains unbroken (gattoplex, 1up 4 up etc)
Deconstructed juggling - continuous breaks in the links (grace, georgian shuffle etc)
Note: In my neck of the woods, deconstructed juggling has been known as “broken down juggling”. I am not sure how wide spread that term is. Ben Richter used the term deconstructed, when he critiqued clay motion, and I thought his post contained a few important observations. Perhaps it is a better term than broken down.