"downtime" is kind of a weird way to call it. Its not down in the same way we think of AWS being down. I guess the right way to say it is 'time when the current longest chain was not an active chain', but that doesn't exactly roll of the tongue I guess.
At any given moment there are a handful of active chains on the bitcoin network. really a better word I should use is fork, because these chains are 99.999+% identical, its just the last few blocks that may differ. The consensus work that Blockchain does is all about figuring out which one of these forks is the one we all agree on. This is why if you deal with actually posting transactions to the bitcoin network they say you should wait for ~6 confirmation blocks to ensure that you don't end up on on a fork that doesn't end up in the consensus (its unusual, but it happens).
During those downtimes the current active chain that is run today was not one of the active chains in consideration for consensus. This is why the fixes was to go back and do a softfork from an earlier state.
