My experience with PfR doesn't include PfRv3 and/or IWAN, but in the earlier OER/PfR, I don't recall the technology supporting what I understand you're asking for, i.e. wait until any link hits 100% and shift excess over to a less utilized link.
What the technology did was proportionally move flows about to try to obtain equal loading on all links. I.e. The goal was equal percentage loads on all links. (BTW, the load percentage was based on link capacity, so, for example, if you had a FE and GE link, each might be loaded at 20%, but the GE would be carrying 10x as much.)
I vaguely recall (???) there was an option for "$ cost" consideration, i.e. if one link charged more for bandwidth usage, it would only be loaded when lower cost links were 100%.
Again, though, waiting for any link to hit 100% before rebalancing, excluding possible bandwidth costs, wasn't done, as I recall, in the earlier versions.
Also, BTW, this technology didn't immediately move flows about, because it sought to maintain a load average over some time period and it also recognized that a flow moved to another link might adjust its own bandwidth usage, based on bandwidth availability. I.e. a kind of dampening to link utilizations. (This "dampening" is important, because actually, any link has only two usage levels, 0 and 100%. When we say something like 50% usage, what is really happening is 100% usage during half the measured time period. So, PfR, by default [?], IMO, works better when it doesn't LB waiting until 100% happens.)
