While not cross-platform (meaning that there
is special preparation needed to run it outside Windows), it is possible to run SuperMemo for Windows (desktop-only) on macOS and Linux via Wine. The Winetricks verb I used[1] works for Linux and used to work unmodified for macOS until (I think) Mojave. In the next days an updated version, with more features, will be packaged for Lutris[2] (which for now, remains in testing phase), making it easier to install on Linux. Since Catalina's removal of 32bit support I have stopped tracking the macOS situation, but I believe it is still possible to run depending on whose patched Wine package you pick up for the task. I hope to try again once I can run a Catalina VM or a kind person is willing to lend me remote desktop minutes.
[1]: https://github.com/alessivs/supermemo-wine
[2]: https://lutris.net/
Failing that, there's obviously virtualization.
> Anki has plugins...
I could dive and swim in the sea of Anki plugins. Though if you look closer, while activity around the extension ecosystem is vibrant, there is something more fundamental in the core of Anki that is producing excessive workloads, and making it generally not up to the demands of a large and varied body of knowledge and vehement use.
One example is the recommendation that a whole body of study is split into separate decks–which, I assume, seeks to avoid biasing scheduling towards that required by the more representative portion of material present in reviews. But you are equipped with one brain and live by a single time line, taking frequent local decisions regarding learning material which have global implications for your performance. When you split your study material into separate units of analysis, these details are too easy to ignore.
In contrast, SuperMemo's algorithms play well with (it is the recommendation, actually:) throwing everything into a single collection (deck)–items of knowledge which you can optionally prioritize individually or in groups–such that the global implications of your local choices are accounted for, maintaining the forgetting index premise, and gracefully adapting workloads to your capacity.
Instead of improving the core functionality in this direction, Anki community's response has been to give you more knobs (scheduler this, scheduler that, load balancer gobbledygook) around essentially the same historical limitations.