If the attacker forms a chain with higher chainweight, checkpoints won't hinder or help the reorg. That higher weight chain will replace all instances of the current chain, regardless of if they have checkpoints enforced or not.
In this scenario, a long-range double-spend attacker would always be successful (and checkpoints would only be helpful for short-range forks),
As far as I have understand Vitalik Buterin's "Weak Subjectivity" paradigm, this behaviour (the chain with the highest weight winning every time, even if it's a fake chain) should be avoided at all costs, because even if the attack is difficult and expensive, there is an incentive to try it if you always will be successful.
Chain-weight, unfortunately in PoS coins is not objective, because it can be faked. That's why some pure PoS coins prohibit long reorgs (the "rolling checkpoints" I mentioned before) and use techniques like Economic Clustering to know if well-known nodes like exchanges are on the same chain like you. With these mechanisms it can be avoided that the attacker "establishes" his attack chain, because most nodes won't follow him even if his chain-weight is higher. But yes, you sacrifice "objectivity" - but at the same time, you dis-incentive these attacks, so they become almost impossible.