Allocation engine does not properly handle exhausted pools for FLQ

Suppose you have a shared network with two subnets. One subnets uses an FLQ allocator, and another uses some other allocator. When the first subnet is exhausted the allocator returns a zero IP address. At this point, the allocation engine should resign from trying to allocate an address from this subnet and move to another subnet. However, it continues and offers a zero address.