Valid concerns.
> whether this could ever practically roll out while simultaneously shutting down the existing welfare/disability/etc systems
Yang's plan was for UBI to be opt-in, in exchange for waiving access to many existing programs (food stamps, etc). If he had gotten traction beyond 5%, we could've expected major pushback from the left on this; similar to M4A, in many ways it's more helpful to the lower middle class than the working class. But I suspect many would still vastly prefer check in hand, both due to the complex bureaucracy (and anxiety) of the existing system, and the elimination of poverty traps.
I think this is a reasonable way to shrink spending on existing "entitlements" without cutting them altogether; and would not only offset the cost, but grease the wheels of political viability (in a case of strange bedfellows, it aligns with the Paul Ryan / Steve Bannon "deconstruct the administrative state" playbook).
> The whole "we'll just tax the 1%" hand-wavy stuff is not a good enough answer here.
True: we'd have to raise more in tax revenue, full stop. I'd like to see a lot more enforcement and closing of loopholes within our existing tax code (Apple parking its cash in Ireland, etc). Easier said than done, I know. That aside, I think a micro-tax on high-frequency stock trades, and/or Yang's VAT, are both reasonable approaches.
My favorite strategy for raising public revenue, and wealth redistribution in particular, is the Pigovian Tax [0], which has an ancillary (primary?) benefit of setting a price on a negative externality. This is the proposed solution from "The Largest Public Statement of Economists in History" [1] on climate change; a sufficient carbon dividend could likely pay for a massive portion of UBI, though it would have to be ramped up gradually. I think there's a strong case for applying the concept in other domains as well (single-use plastics, for instance).
Finally, it's worth considering that all the hand-wavy trickle-down logic of "supply-side economics" might actually work when pointed in the other direction [2], like a stimulus package that never stops. In addition to reducing silent inefficiencies of economic struggle (addiction, short-term thinking, anxiety-induced low executive function), most of those receiving the dividend would spend it, increasing economic activity and therefore tax revenue. (There are potential problems with this model, the biggest being the risk of the lion's share going to landlords; but that opens up a whole other rabbit hole of Land Value Tax, Henry George, etc.; we arguably have a rent-seeking problem in that domain already, inducing a hidden tax on the working class with or without UBI.)
[0] https://en.wikipedia.org/wiki/Pigovian_tax
[1] https://clcouncil.org/economists-statement/
[2] https://en.wikipedia.org/wiki/Demand-side_economics