@seattlecyclone, thank you for the extensive reply and comments. I'm not going to quote it because it's rather long, but wanted to reply here with some of my own comments. OP, if you want me to start a different thread or take it to PMs, let me know.
I agree with your point about a side gig possibly being better to MAGI than investment withdrawals. In my actual case, I have side gigs, capital gains, interest, dividends, and Roth conversions.
The actual formula for the ACA phaseout ranges can be found if you poke around the web long enough. IIRC it's three line segments.
Regarding marginal rates and conclusions, I didn't spell out my scenario precisely before. I fund my living expenses first from any taxable income and non-taxable refunds, gifts, etc, then I sell from my taxable, which generates some LTCG. Beyond that, I prefer Roth conversions over additional voluntary LTCG based on some reading I had done elsewhere and the fact that my taxable is relatively small and my tax torpedo at age 70 is likely to be large.
What I did last year was to look at varying amounts of Roth conversion and pick the amount that absorbed all of my CTC, kept me in SNT range for the FAFSA, and was at a total marginal rate that seemed reasonable. Between federal, state, ACA, and EFC, I ended up in a 28% marginal bracket.
What I think I should do this year is the same, except I probably should be willing to be in a higher overall bracket, because I was ignoring state income taxes at age 70 when looking at the tax torpedo. If current trends continue, I'll owe 24% marginal and 7% state. So I really should be willing to pay up to about 30% now - if I believe that evening out taxes over the next 20 years (I'm 50) is the way to minimize taxes overall.
I have a fair number of other complexities in my financial situation which make it hard for me to try to predict and therefore optimize, including ACA, kids in college, child support and associated constraints, potential LTCG from a QSBS, a potential inheritance, etc. I have tried to build accurate models of this in Excel before and they collapse under their own weight, so I usually just try to achieve a level of "not the worst option I could have picked". For example, I'm doing some Roth conversions now, which is possibly not as good as somewhat larger Roth conversions, but also probably not as bad as zero conversions. I may try to build a model in Excel again sometime soon.
Thanks for the correction about the dependency exemption. You are right; I actually meant the CTC/DCTC.
Thanks for pointing out the 9.86% range. I'll have to look at that more closely when I have more flexibility in my health insurance choices. Currently I'm locked into a CSR87 silver plan until my youngest graduates from child support.
The fortunate thing is that in my case, as long as I make reasonable choices, I'll probably have success, as my safety factors are very wide. But I also get frustrated that I can't seem to get my optimization completely built out the way I envision.
Thanks again.