That’s a big question.
As far as stability goes, and as far as openness goes, Ada is top-tier: I’ve compiled 30 y/o code with a compiler that didn’t exist on an OS that didn’t exist, having to do two things: (1) change a dozen or so instances of an Identifier that had become a keyword, and (2) split a single file because GNAT has to have spec/body file-separation. — Trying the same thing with C is an exercise in patience. — So, insofar as that goes, Ada is top-tier.
AdaCore doing Rust is only slightly relevant: AdaCore is a company, not the language. While thy have been a large contributor, the language itself is freely available, open to anyone who wants to implement it. — That leads us into the fact that the LRM is the same as the ISO standard (modulo formatting/templating the ISO requires).
Perhaps the biggest thing that would be helpful to the language, at least my opinion, would be the addition of a meta-language feature allowing (a) the definition of what is essentially Abstract Data Types, which would allow you to [statically] hang the annotation for SPARK proving there; and (b) the definition of how a type is used/interfaced [this would clean up user-defined indexing, as well as allow things like all types to have finalization], essentially expanding the generic-formal-parameter notions into the language, allowing us to clean up the language. — This would allow a lot to be obsoleted, simplifying the language, and the ability to define the language within the [meta-]language, allowing the standard to be proven/model-checked, in-turn allowing removal of some of the trickier verbiage.
Ada is a good investment.
Even if you don’t use it professionally, it will alter how you think about programming in a good way.