1) The assumption of reliable time-bounded propagation. This can be disturbed in various ways by the attacker.
Counterpoint: If the attacker already has this type of power, he doesn't need anywhere near 51% to control the network. Compare bitcoin with it's 10-20 nodes with the network view. An attacker need only disrupt a couple of them to cause complete chaos. In Decrits, he will need to disrupt massive portions of the internet.
2) Race conditions with the timeouts. The malicious peers can release the critical information when its time is running out, in this way dividing the opinions of the observing nodes. Some of them will believe the message is late, others that it was on time. This confusion potentially will open the possibilities for other attacks.
You are going to have to be clearer. I can imagine several different things that you mean by "timeouts" so instead of covering them all in a jumbled mess, I'd prefer clarification.
3) The nodes which didn't observe the attack, and only see the resulting fork, will still not know whom to trust. Both parties will claim that the other didn't release their TBs on time or didn't accept legitimate TBs which were released on time.
I explained why this is ok--I'm sure not very well though. A PTB will effectively stop the attack on the network, meaning there will not be any 51% attack, but just an attack on perhaps 5-20 honest SHs (the attack you and anonymint brought up as potentially being more critical--making small forks). But those SHs that were attacked can not be denied consensus, because the evil chain must include the PTB or it is screaming network takeover attempt. A PTB should essentially be thought of as a critical stop to network activity. Until a node sees a chain with it included, it should not transmit that chain, and and anyone using the network should not make transactions until they see a chain with that PTB.
It sort of puts EvilCorp on the spot. Stop the shenanigans or go ahead and fork. Everyone that is currently watching the network will see the fork forming. It won't happen instantly, it has to be as a factor of time. If they continue to fork, more and more people will see the untrustworthy network for what it is. Everyone watching knows of the attack. More will start asking for TB data, more will start seeing the fork.
Even if people are late to the game, they will receive the missing TBs assuming they are not isolated. To actually destroy the shares of the honest network, it must pretend as though the honest SHs refuse to reach consensus. While a TB from 2 hours ago could have been made 10 seconds ago, a new node will keep a counter from the time *it* saw the TB. Honest SHs must still add section 1 to the chain regardless of how late it is (unless it is so late that it is past the absolute deadline where shares are destroyed). This is what brings consensus. So anyone hopping on prior to that deadline will start counters. If there is a fork, the honest half will have all, or mostly all of the signatures in consensus that any node could see (including but not accepting). The dishonest half must exclude those whose shares it wishes to destroy.
If it does not exclude those, they only receive a strike. It is an attack vector, but it is not a critical one as I said. If they include the PTB then start
another attack, the untrustworthiness will build more (it would have to go down over time). So if they do not want people to know who all of the dishonest SHs are and provide a long period of time for everyone to find out, they must settle for giving a few, random honest SHs strikes.
I really don't think this can be countered by any "well an attacker can control this or that view of the network". If they have the capability to do something like that, any defense for anything is impossible. It is essentially saying "you have fixed every single thing but complete and utter control. Your currency is 99% attack proof." And I say "thanks, that's what I been sayin brah"