We’ve moved this ask to the backlog due to technical challenges in reserving Public IP addresses during stop state.p l supported this idea ·p l commented
Folks, this is most likely because somewhere in the service infrastructure there is a cost per connection licensing fee. Might be to an outside hardware vendor, or the way the underlying network infrastructure cost is calculated. It's probably the fixed cost of each connection that is expensive, and it has nothing to do with bandwidth, CPU, etc.
This generally happens when a product has to be built and provided quickly. You rely on outside vendors or other internal teams to help you build the service and they nickle and dime you with per item connection costs or you feel you need to recover the initial investment in something.
Or, I guess it could be something even dumber. Maybe there's a metric or goal around a fixed number of $$'s per connection. And this would be the only way to get there.
Or something completely different. But either way, it's clearly not been designed as a grow on demand, fully scale-able, kind of service.
Whatever it is, it's clearly a "cut your nose off to spite your face" short sighted kind of mentality. And you won't get a fully scale-able service until they fix the real underlying problem.