OpenFaaS has a rest API, so no that's not accurate - you can scale to zero and back again.

You may have been "toying" - many people do, but here's some examples of end users https://www.github.com/openfaas/faas/tree/master/ADOPTERS.md

faasd uses pausing of containers so there's barely a noticeable cold start unlike with Kubernetes where it can take a couple of seconds.

Up until 2018 everyone just used a min scale of 1, and was very happy because it meant never having to bear a Kubernetes Pod being scheduled. This is one of the reasons that faasd is quicker - no distributed state to manage.

Feel free to ask for help on Slack if you are unclear about scaling. I'm sure the community would be happy to help.