Evolution of difficulty on testnet3

[mriou]

Hi,

I'm trying to implement the difficulty calculation algo and test it on testnet3. But I can't explain how it seems to go down even though ranges of 2016 blocks (a block every 10 min for 2 weeks) are added in much less than 2 weeks. For example between block 6048 and 8064:

http://blockexplorer.com/testnet/block/000000000be66197ad285aedd52e56036f28d595fe281858bc5d562173d4d6de
http://blockexplorer.com/testnet/block/00000000118da1e2165a19307b86f87eba814845e8a0f99734dce279ca3fb029

There's only a little over 5 min between those 2016 blocks so the difficulty should have gone up by the maximum factor of 4 from 16 to 64. However it decreased instead from 16 back to 4.

Could someone help me understand this? I've looked everywhere including the rules for testnet3 and the difficulty calculations as well as the source code but nothing seems to explain this behavior.

Thanks a ton!

[1714860756]

1714860756

[1714860756]

1714860756

[1714860756]

1714860756

[johoe]

I guess this is because of the fake time stamp in block 8063, which is two months in the future.  It seems that someone wanted to test exactly this behaviour.

[gmaxwell]

You're welcome.