Lambda count "drain" workers as "ready for scheduling"

[MaxymVlasov]

How it works now

1. Let's imagine a situation when Workers: 0, Busy: 0, Queue: 0
2. Then, some run triggered and we get Workers: 0, Busy: 0, Queue: 1
3. Lambda sees that Queue > Workers, create a new Worker, and run scheduled to it. We get Workers: 1, Busy: 1, Queue: 1
4. After the run was completed, we had Workers: 1, Busy: 0, Queue: 0 for some time.
5. Lambda sees that Workers > Busy && Workers > Queue so it starts termination of Worker drains it.

So far so good.

6. After the Worker drain, there is another run pop-up, so we have Workers: 1, Busy: 0, Queue: 1

BUT, because Worker is drained and Workers is not > Queue, run in Queue end in long awaiting, till old one Worker will be terminated, and Lambda detect that Queue > Workers and create a new worked (step 3):

Workers: 1, Busy: 0, Queue: 1 -> Workers: 0, Busy: 0, Queue: 1 -> Workers: 1, Busy: 1, Queue: 1

Suggestion

In Lambda, Check the status of workers during the Queue > Workers evaluation.

Currently it is Queue > all Workers, when should be Queue > not drain Workers