This mining group have 51% of the hashrate. If they remain coordinated, i.e. only mine their OFAC-compliant blocks *and mine only on top of each other's blocks*, then they will have the longest chain
When they first decide to do this, then it's possible that - for the first couple of blocks immediately after they 'activate' - the 49% might be lucky and will build up a small lead. But that's just temporary during the transition. The 51% will keep building their chain, ignoring the non-compliant blocks from the 49%, and the 51% can't be stopped by the nodes or by the 49%.
The nodes have zero defence against this kind of coordinated action by 51%. The only thing we can do is try to set up the incentives such that the miners remain uncoordinated and do not attempt this.
This is why is remains critically important for the relay network to remain the primary (ideally, *exclusive*) source for the miners to get their transactions. If we try to censor things, encouraging the miners to further extend their "private mempools" with their out-of-band transaction-submission systems, then the miners will build more of those kinds of systems and eventually the *big* miners will connect their private mempools to each other. Eventually, the *big* miners will realise that they are simply coordinating transactions with each other and that they don't need the relay network and the relay network will be less and less important
In that scenario, the relay network will simply be passive consumers of the blocks produced by the big miners; the transactions will be decided exclusively by the *big* miners. Bitcoin will be dead.
The relay network must be optimized to keep the miners divided, and to keep all the miners, especially the big miners, dependant on the open relay network. As long as we keep the big miners divided from each other, we noderunners remain powerful. But we'll lose that power if we don't relentlessly continue our 'divide and conquer' against the big miners