Or, to quote Ernest Hemingway, "Gradually, then suddenly."

If this was flu, say, with a smaller k then you'd see a more normal exponential increase. With this it probably looked more like 1 person, 1 person, 1 person, 1 person, 1 person, 2 people, 30 people. Even absent protective measures most people who get Covid transmit to 1 person or nobody. But some spread it to dozens of others and a small number to hundreds.

I'd guess this is mostly driven by the orders of magnitude difference in peak viral load we see with SARS-CoV-2? But that's really something that needs research.

The 70% is my issue with the article as well. That seems to indicate a 30% chance for the pandemic that actually happened, but that seems awfully high.

30% chance for a pandemic not necessarily this pandemic. Assuming they are correct the outbreak could have easily been noticed in a different city for example.

Aren't there other outcomes, like the relatively contained SARS/MERS outbreaks of the past?

Even if the outcome isn't particularly rare, I still think any bad outcome from a random process counts as unlucky.

I suppose you're right, but if you're paying Russian Roulette with 5 out of 6 bullets loaded I'm going to be more predisposed to thinking you were playing a really stupid game. I don't consider my lottery losses every time I buy a ticket unlucky, I just consider them the expected result.