Mr. Green, now that I'm digging in to this, I feel like there are more people doing this than you'd think. And I'm now contemplating doing a big (like $100K) Roth conversion this year, while I'm still working. Have you spent much time with i-orp? I know your situtation is different because you have real estate, but I think it might be powerful enough to handle it. And there are paid options, too, though consensus seems to be they aren't really better, and in some cases are worse, than i-orp. The main problem with i-orp is that the brilliant, kind soul who runs it is in his/her 80s, and unless someone steps up with a succession plan and agrees to take it over, well...
I found several interesting threads by searching for "i-orp" in bogleneads. A real rabbit hole!
ETA: the Bogleheads have already helped me a LOT. One reason i-orp is so aggressive with conversions is because you tell it what your current asset allocation is in each account--and if you are already doing the tax efficient thing, your Roths have all equities and all your bonds are in your tax deferreds. So to optimize, the program wants to get you out of the lower performing account (the tax deferred) and into the high growth one (Roth). But that, obviously, completely screws up your asset allocation (unless you are 100% stocks or 100% bonds, but most of us aren't). To "fix" the problem, you lie to i-orp and tell it that every account (Roth, tax deferred, and after tax) all have the same asset allocation. Once I did that, it no longer recommended I do conversions up to the 24% bracket and do them fast, but just recommended up to the 12% and do them slow.
So this "solution" isn't perfect--your Roths *should* be more aggressive, and will continue to be your whole life. So i-orp's output will assume a (slightly?) worse outcome for you than you'd actually experience, following its advice. But anyway, it's still an incredibly powerful tool.
Note, too, that the wisdom of aggressive conversions is a function of predicted returns of both equities and bonds. It's *really* sensitive to even a percentage difference in prediction. So I don't know what to make of this. Possibly that I should just go somewhere between "super aggressive conversions" and "no conversions." But man, this stuff is hard.