This reduces the security model of the entire chain to that of POS. Any attacker can just wait for a POS block, then use his stake to perform a double spend by minting a block and just keep doing that forever at zero cost every other block.
This is incorrect for several reasons.
1) The scenario you described is only valid for the POS block being currently mined, however this is no different than POW miners attempting to mine blocks and come up with 2 blocks around the same time. The network will eventually go with one fork and the other will be orphaned. In this case, it is the POW miners who decide which POS block is built atop.
2) Yes, you are right that pure POS is insecure because of the grinding problem which pretty much every POS coin tries to resolve via checkpointing by trusted parties.
In Interleaved mining, the checkpointing is instead done by POW miners based on whoever mines the next POW block just as the correct chain history is decided by POW today, POW will decide which POS block is the preceeding block.
To simplify this, think of POS blocks no different than ordinary Bitcoin txns, POW miners decide which is valid and in what order.
3) While i have not described the POS algo in detail yet (wish to get the high level idea across first), my current thinking wrt the "strength" of a POS block would be a combination of coins * age where inclusion of a POS block will consume coin age, thereby greatly reducing the probability that the same party will mine another POS block in the near future.