it happens...

ultimately it's random but in experiments i've done with various methods i've found there are some ways to tighten the variance...

the first method i found that worked was having multiple different block "types" that each had to be distributed according to a different distribution, 9 of those with a doubling of duration between them narrowed it down to about 50%, ie for bitcoin that would be 5/20 minutes, for more than like 99% of the time

a second method had to do with using a more advanced type of adjustment, which adds the integral of the error rate (variance versus target), and though i didn't get it right at the time, the derivative of the base variation curve, if it has a high frequency squelch applied to it (reducing the high frequency component of the derivative) probably also can get to the range of the 9 way multi-schedule scheme

i don't know what the probability of 1 hour between blocks is, but it's maybe less than once a year, so maybe it is a lucky event

the halving also fell right as a CME hit the earth btw, did you know that?