Re:frontend, you're absolutely correct. Magento's frontend is a sad state. MageOS is actively looking for ways to move Luma/Blank into a separate installable package so that devs are not forced to see it at every turn.
In terms of Headless, there are a bunch of different heads at this point, but none have yet solved what I call the "upgrade-safe theme" problem.
I like what we've done with Daffodil since it doesn't impose any of those limitations that come with themes, while still handling a lot of the complexity that devs face when building ecommerce stores.
If you can make a name for yourself, you can do very well. Especially for Magento, I love the fact that I can make small change, deliver it as a composer package, and then (if the code is good enough) see the results on many brands within a few months.
It's hard to get that close-to-customer visibility anywhere else as most "cool" large scale FOSS projects are dev focused, which means your impact on the lay user is less obvious.
The backwards compatibility is really a challenge, but I think more importantly dumping Knockout and Luma and all that insanity and going towards the Hyva route, or going more-headlessy with a good starter theme would encourage people enourmously.
Magento's frontend cluster-F is the reason we moved away from it. It was great to have such a good pricing engine, integrations pieces, currency support etc, but all of that skillset which is magento's core capability being a really well rounded online OMS is overshadowed by long and expensive turn arounds on frontend work.