There is a logical possibility that the core functionality of the Earth system is to handle any kind of destabilizing stress, be it an asteroid impact, or over-expansion of some new highly effective but greedily species. Yeah, I am talking about Earth as a functional cybernetic system - the Earth "knows" its forces, it can order them to solve strained situations. The Earth might have done it many times locally, and on several occasions globally. Nothing happens by mere deterministic chaos... Damn, even seismic events might be linked cybernetically to global warming, as they are powerful signals to biosphere. The Earth might already be "committed" to a course of action... but it might be following and learning signals of our behaviour as well. If this interpretation is remotely true, our technological solutions might be marginal compared to unleached Earth's powers. We would be facing the challenge of surviving the rewind of a geological cycle, provoked by ourselves. On the other hand, the cybernetic model offers the possibility that the Earth might "appreciate" measures of genuine stress relief, that humanity and the Earth might communicate for mutual benefit, ha ha.
I leave this as far fetched as it might seem.
...and therefore we shouldn't be doing anything? That's not a rational decision.
Assuming there are only two options (do nothing or put up an SO2 sunscreen), and both options are poorly understood, then it is still the rational thing to go through with the option where the better outcome is predicted. No matter how large the uncertainties are.
In the present case, however, there's no conflict between expectation and risk. As best I understand it, the negative effects of unchecked warming also produce what seem to be far greater risks -- setting positive feedback loops in motion, changing weather patterns, possibly switching off the Gulf Stream... Words and ideas I offer here may be used freely and without attribution.
As it happens, we already know the consequences (smog, acid rain, global temperature drop if scale is large enough) of a massive SO2 release, as you have pointed out. So the modeling should not be too fraught with uncertainties.
Carbon sequestration, which is on everyone's lips these days, is a most likely a problem with more uncertainties--how carbon dioxide will affect subterranean rocks and soils; question of leakage and release after a few hundred years, etc.
