Something with a cutting-edge game engine like Bevy: Entity-Component-System architecture, Rust, immensely multi threaded, new graphic tecnhiques like Meshlets (same as nanite tech from UE5, the other only game engine I know that has it), physically based rendering, highly performing, customizable, with good multiplayer capabilities, using new techniques of software engineering (it's not the 90s anymore).
Something with a community that embraces collaboration and the new tools (again, it's not the 90s anymore). Git forges, chat platforms, RFCs.
Something that from the game engine to the flight models is open to be reused across academia, different types of sims such as land vehicle sims, civil aviation sims, combat sims. Something that wants to foster that kind of relationships.

All of this is possible, but not particularly easy. It doesn't need to start big, just with libraries and utilities for academia and developers that one can build on top of.

Hence why I think the formula is "Bevy + Blender + some Copyleft licensed parts (GPL) + Community". I've given quite the thought to the topic, and a custom ECS engine is paramount. One that is designed for working collaboratively and not by in-house devs with UI tools for it. One that is massive in the cutting edge of tech and at the same time easy to collaborate on remotely. That is Bevy: the shortcomings (no UI tooling l for now) don't matter for Sim games, as we normally need just one model in Blender, rigid animations, and no level editors. It also is written in Rust, is performant, a bliss to work on iteratively and grow the size of, and people are actually looking forward to work on for free, contrary to C++.

Whatever we do, the best we can do with fellow enthusiasts is recognise that sim games are a captive market. This way we can change the Zeitgeist, and move away our attention from the hype and drama of this companies (Microsoft, DCS's eagle dynamics, IL2's 1C), and into collaboration.