Confused out autoscaling deployment sleeping

I’m very confused about deployments vs “development” AKA clicking Run in my project…

One person on the forums said

Autoscale deployments will not go to sleep, even if very few requests are are fired.
Autoscale and deployment questions - #4 by IroncladDev

And someone else said

Autoscale deployments go to sleep between requests (referring to direct HTTP requests such as browser access)
Deployment / Always On - #3 by lincoln-replit

So which is it? I’ve been playing around with clicking “Run” vs deploying an app today and I’ve somehow already burned through 6,420 compute units in just a few hours of testing. I’m boosting my app which is free with the Hacker plan…but is that actually just chewing through my compute units? I’m very confused by all of this.

Additionally, how does Boosting differ from Autoscale?

Answering half of my above post - “Run” will turn from into starting in Jan and will no longer be able to spin up on demand, but turn into a published live dev environment as long as someone is actively in the replit project. It does sound like autoscale will go to sleep…or go to zero resource/compute unit spend when resources aren’t being called, so not sure why there’s conflicting information about that. Autoscale won’t be able to run background workers or anything a typical server would do like persisting data - VM is needed for that.

Haven’t found an answer yet to the Boosting vs Autoscale. Is Boosting just for the dev environment for and and does not spend compute units with the Hacker plan?