There's likely a decade+ tail after the end of active particulate injection. If there is too much to begin with, you piss off all the electorates with years of chaotic weather, threatening food production, financial markets, and ecologies all over the world. Get it just right, you wreak less havoc, piss off fewer electorates, and lower global average temperatures. Get too small an effect, you risk overcorrection or the project failing before it has a chance to really work.

There's also the fact that you're altering a chaotic system, with fundamentally unpredictable side effects, some of which could last longer than a human lifespan. Wanna make California get monthly hurricanes? How about weekly tornados in Colorado? Bury New York city in ice?

It's not that any of those things are probable just that they're possible, and chaotic systems can drastically, catastrophically, and exponentially change configuration, and no matter how good your supercomputer simulations are, you fundamentally cannot predict the consequences.

All of human civilization teeters on the current climate configuration - geoengineering has terrifying potential. Humans have a horrible track record in attempts to predictably alter chaotic systems and ecologies. We damn well better know what we're doing before we use globally acting tools.