Other sources report the only 'days' delay was between block 0 and block 1 – and thus likely an idiosyncratic effect of the launch. Apparently, also, there were a few more >20hr blocks in the 1st half of 2009, when interest was very, very low.
So while technically true, without further details, your comment implies something that's not really reflective of the system today.
Of course, as a stochastic process, it's possible that the delay reaches any length whatsover. But it becomes vanishingly unlikely, under the rough assumption that hash power stays at the level observed at the last difficulty-adjustment.
Anyone arguing a delay between block 0 and block 1 is not arguing from a rational standpoint. It is akin to arguing that an airline had days of no flights where they didn't fly any routes at all, because the company was obviously founded and incorporated at least a few weeks before their first commercial service was introduced.
So while technically true, without further details, your comment implies something that's not really reflective of the system today.
Of course, as a stochastic process, it's possible that the delay reaches any length whatsover. But it becomes vanishingly unlikely, under the rough assumption that hash power stays at the level observed at the last difficulty-adjustment.
By the formula Tadge Dryja shared, a 1-day (1440 minute) delay would be expected once every 6.568×10^57 average Gregorian years, or far far far longer than the age of the universe so far: https://www.wolframalpha.com/input?i=%281%2Fe%5E%28-1440%2F1...
(If large amounts of hash power disappear for any reason – war, regulations, sabotage, bugs, massive price drops, etc – then all bets are off.)