There's never been a very strong rebuttal to my observation that: Downvote strength only need to be sufficient to push abuse posts rewards towards roughly the new curation value in order to deter them rather than all the way down to 0.

I don't agree this is sufficient for what we actually want to accomplish. I understand the house is burning but let's not lose sight of the actual goal beyond putting out the fire. It is a necessary condition to avoid total failure (as we have today) that self-voting not be more profitable than curating. However, that accomplished, there are still situations when people, for whatever foolish, malicious, ignorant, mischievous, competitive, etc. reasons, want to spend money which other stakeholders do not agree benefits Steem. This is bad. It is not the same kind of bad as self-voting, but it is still bad.

We want rewards pushed toward the genuine contributions. We don't want rewards pushed toward pictures of dog poop because the voter thinks it is funny doing so, and doing so has little cost to them. You, with your statement above, want to make voters indifferent (or perhaps slightly tilted toward not doing it) to self-voting. I want to make them worse-than-indifferent to any sort of bad/nonconsensus voting, even if not directly profit motivated. The former is a subset of the latter, thus the latter requires more available downvote power. We need to push bad voting not just to the point of indifference but meaningfully below it, to ensure that people looking for self-amusement or to cause damage don't view an infinitesimal cost as worth paying. The assumption (always questionable anyway) of fully-informed hyper-rational profit-motivated actors is a not sufficient vehicle to protect the community treasury in my view.

Let's set that aside for now though and consider your above quote again:

Downvote strength only need to be sufficient to push abuse posts rewards towards roughly the new curation value in order to deter them rather than all the way down to 0

This is exactly the very strong rebuttal (which you claim doesn't exist) of your 10% number that I made in my reply above. Self-voting is (nearly) 100%. The new (proposed) curation value is 50%. So by your criteria we are looking to push it down 50%. I have further pointed out that curation has an added cost (passive/dumb voter considering an alternative of curation will need to either invest effort or split rewards with a curator). This means the push needs to be "something" more than 50%.

Now we can speculate on all sorts of things like how much of the available downvote power will actually be used, how much fear of retaliation will remain, how many voters will attempt to get away with self-voting and need to be pushed back, how much dispersed downvote power can be concentrated on some assumed number of abusers, etc, etc. but as a first order approximation without relying on favorable assumptions or wishful thinking, this is the number/range we are looking at: something above 50%. To my view that is close enough to 100% to just KISS. It is quite obvious that not all downvote power will be used just as today not all upvote power is used despite a clear and direct (and orders of magnitude larger) monetary incentive to do so. So what is the exact number (for max sufficient available downvote power)? I don't know, but it isn't 10%.

Ok, so an actual separate pool is pretty much out of the question due to implementation limitations

I didn't exactly say that. I said it was more complicated than individual downvoting power. It might still be plausible to implement. We'll have to see when actual coding is done. I would not reject code that took either route.

would you agree to 50%?

Hell, I already said multiple times (including the chart above) that I would agree to 10% as a step in the right directly and clearly preferable to the status quo. Sure I would agree to 50% too. But I don't think it is ideal. We won't get ten bites at this hard forking apple to try things and try again when it doesn't work. I want them (in fact, more likely, it, at best) to count. My preference is to really drain the swamp, not try half measures that might, in some idealized model, barely work on the margin. Let's Make Steem Great Again. lulz.

and with a half ass coding job

We don't want that and I wouldn't support it regardless of the merits of the attempted change. I don't think that's really a viable path at all.

steemit's help on any of this is out of the question until well pass SMTs

That appears to be the case. They seem like they might be willing to do minimal things like some code review on well-presented already-complete work but not more than that. This (as anything) could change but it is the current working reality.