I found it a bit unintuitive that three blocks in a row at an average of 10x the expected rate is so much less unlikely than a single block at one 10th the expected rate. But the math seems to check out. And if you look at the graphs from https://blog.lopp.net/bitcoin-block-time-variance/, you can see that 1 minute is actually around the peak of the historical time delta distribution.

poisson.cdf(k, rate) is the cumulative Poisson distribution. I.e., it returns the probability that the number of events within an interval is less than or equal to k, with rate being the expected number of events.