Although as a counterargument, one would expect the decline in hashrate would have happened much faster if it was coordinated.
Theoretically speaking:
It takes time to reconfigure or even factory reset machines. It did happen fast, in a few weeks/epoch. I think that's as fast as it gets if you talk about millions of machines.
Over exaggeration always useful to help someone think their own FUD is true
If we use the term stated "few weeks" and make it 8 weeks,
8 weeks = 4 diff changes (again we'll exaggerate that also since it was a less than 4)
So time wise we are saying since 13-May since the -15.97% diff change started from the previous diff change.
So actually ... 9.4 weeks ... wow amazing how a '
few weeks' can suddenly mean
9.4 weeks ...
4 diff changes = -15.97% -5.30% -27.94 -4.81%
= 25046487590083.3 down to 13672594272814.1 = total 45.4% diff drop
~13.4T drop = ~81.4EH/s
If you go 2 diff changes (4.9 weeks - maybe you could over exaggerate few to be 4.9?) it's -27.94 -4.81%
= 19932791027262.7 down to 13672594272814.1 = total 31.4% diff drop
~6.3T drop = ~38.3EH/s
So I guess we must have two different convert mining attacks?
The first ~half switching off machines more than a few (4.9) weeks ago and the second half switching off machine over the last few (4.9) weeks?
Numbers suggest the hash rate start has started to increase towards the end of this diff change that just happened.
I guess they forgot to enable the attack when they switched on some machines? ...