If I understand the whole system correctly, the difficulty is adjusted **every two weeks** adjusting it in a way, that the blocktime stays at ~10mins. This means that a sudden loss of a large amount of global hashing power would lead to considerably higher blocktimes until the next difficulty adjustment.

It is only adjusted every 2 weeks if the blocks are happening on average every 10 minutes.

The ACTUAL trigger for the adjustment is that the block height be a multiple of 2016. If the blocks are happening on average every 10 minutes, then that is:

2016 blocks X 10 minutes = 20160 minutes

20160 minutes / 1440 minutes per day = 14 days = 2 weeks.

However, lets imagine that 75% of the hash power were to just disappear immediately after a difficulty adjustment. Since only one-fourth of the original amount of hash power exists, the blocks will take 4 times as long, or 40 minutes on average. This would continue for 8 weeks, before the next adjustment would bring the average time back to 10 minutes per block.

Additionally, there is a limitation on the difficulty adjustment that prevents it from adjusting by more than a factor of 4.

Therefore, if more than 75% of the hash power were to just disappear immediately after a difficulty adjustment, the problem would persist even longer. Lets imagine that 93.75% of the hash power were to just disappear immediately after a difficulty adjustment. Since only one-sixteenth of the original amount of hash power exists, the blocks will take 16 times as long, or 1600 minutes on average. This would continue for 32 weeks, before the next adjustment would bring the average time back to 40 minutes per block. This would continue for 8 weeks, before the next adjustment would finally bring the average time back to 10 minutes per block.