I completely agree with you, a language that considers cloud development as a first class citizen must embrace something like open-telemetry for distributed logging (spans, traces, metrics, logs). When you log something or measure something in your code it should always be part of some distributed trace. We want to embrace this thinking for localhost development and allow our customer to choose their APM provider when they deploy. In a sense we don't see a difference between a cloud resource that is called logger (can be implemented by DataDog, New Relic, Coralogix, etc) and a cloud resource that is called a Queue (can be implemented by SQS, Redis, RabbitMq, etc...)
Also, as you mentioned, because in practice all the `inflight` (https://docs.winglang.io/concepts/inflights) code is async by nature, the language should wait for any API call to be resolved (e.g. "await") by default, deferring an API call is the exception
And structured concurrency. Not async/await, promises, and so forth.
--
Along the lines of "we were promised jetpacks", I'm still stuck on 1990s era future perfect notions like software agents and grid computing. Like the cloudlet (vs applet, servlet) manifestation of Sun Computer's Jini and JXTA.