There's a probability problem here that I think confounds a lot of people. That is that given enough time, an unlikely event becomes inevitable. So the question is obviously what's the likelihood of something going wrong in a given year? For even a well-run BSL-4 lab, that's a hotly debated question, and estimates vary by orders of magnitude.

If, say, the probability of a leak occurring is 1/1000, then the chance of a leak occurring in 10 years (assuming equal weights, when in reality the chance probably goes up as equipment ages), is simply 1 - (1 - 1/1000)^10 = 0.01. One percent isn't great, but maybe it's tolerable depending how bad the agent is. But if the odds are 1/100, then the chance of a leak in 10 years is 9.6%, which is really bad.

Now let's suppose the lab isn't being operated at maximum safety, and the odds shift from 99.9% safe to 95% safe. In that case, in 10 years there's a 40% chance of a leak. Basically, we're fucked type of territory. These calculations require 3rd grade math once you know the a priori odds, which are probably unknowable, and therefore should be assumed to be a lot likelier than you think.

I don't know that most people think in those terms, but they should. In the linked article, they're contemplating what would happen if hoof and mouth disease got out and the livestock industry was fucked. Here, we're talking about causing Great Depression II. I just can't believe anyone would be that fucking consciousless.