You are actually right, originally the target times where much lower as we started the stuff on 1 algo at T = 120. Where we now do T = 480.
But I do see this as 2 separate proposals, we have Tip004 and the proposal to reduce the window. I am purposly trying to keep them seperate in the simulations as they both impact the solvetimes, but also change it. They might work against each other. If we simulate them separately, and they both are positive, then we know they are good, and also how good. If we only simulate them together, then we mask the potential issues with one of them.
I will rerun the simulations on the actual network hash estimates, but I think using random input data gives cleaner results as the actual mining on the actual network is random by design.
I tried to get new simulations on the actual hash rate, but I could not get any clean results that show anything different that we have not shown or talked about here.
We talked about having more hard forks in the previous tari show, so perhaps based on the simulations, and all the work zawy lets try 45 and see how its going to go. I have my doubts about the oscillations, but I dont think its going to be deathly bad.
I’ll see about re-adding the standard LWMA simulations (without penalty) back to my little simulator and running some more trials.
But if that’s the case I’m inclined to agree with you. Neither of us think this will be a complete fix for the issue anyway, but a bandaid will give us time to think of a better plan - if it ends up being needed at all.
Intuition suggests that if we were to cut off the third era with a max around block 240,000 that deviation would be more reasonable. Certainly before ~270,000
