Towards a practical and effective "TLS"?

With the absurdity of HTTPS and certificates in mind, is a practical opportunistic transport layer link actual encryption: - possible? - worth pursuing (almost no one controls their hardware - "it's all VMs nowadays")? - if not worth pursuing, what set of attributes of a society ('national') level internet would need to be in place, for 'sane' opportunistic link encryption to make sense (e.g., mesh networks of various forms, physical server in every household, i.e. fully decentralise our networking, and server/host infrastructure)? - if worth pursuing, either now or after some level of attributes as named is in place, how might it look? Thanks, Zenaan