They do - just slowly. The backup in the high priority queue (G+, LL, and DAS users), causes the low priority queue's estimate to increase... because the low priority will wait even longer when the high priority queue backs up. They will increase the ratio the high priority queue gets to try to relieve the backup.
You have it backwards... both LL and G+ allotments are fixed and can be managed within reason. The problem is the uncontrolled DAS demand. Take a simple example... you want to allocate 100 slots during the time window... You have a model of when people return vs their time slots... so you know if you sell X return times, it will result on average Y people showing up in the time window. If you want to plan for 30% your slots to LL, 50% to G+... and 20% to DAS/other.. then you know how many LL and G+ to sell and keep lines managable.
But now, instead of DAS returns taking up their 20 slots... imagine it balloons and they take 100 slots. You can't turn them away like you do when you stop selling G+ and LL. They clog up the system.
What needs to be done for your question is to set DAS return times based on not just standby's wait, but what the high priority queue's commitments are (including DAS return times) during the same period. This could mean DAS usage causing other DAS users to get longer return times than what just standby is estimated at now. That would allow the system to adapt quicker than just tracking to standby waits.