So, this would be a better question for someone who has actually worked with other actor-model frameworks. But, one sense that I get is that in Erlang, there is a design philosophy where most actors are intentionally stateless so that if anything goes wrong they can die and be replaced easily.

So ironically it seems like many people expected statelessness to be a property, which is the opposite of what we were going for!

Disclaimer: I haven't worked with Erlang myself and I'm probably missing some nuance here. My background is in object-capability systems, which also commonly claim to be actor systems, and match what we're doing very closely.