@Peter R, good summary of the issue/argument. A few thoughts:
In the current situation, miners don't need to actually validate blocks. They can just download them, see how they update the UTXO set, and never validate. The difference in incentives under SegWit depends on the relative costs of downloading vs. validating transactions.
If downloading is the main cost, then forcing everyone to download the full block before they can start including transactions results in less incentive to skip validation, since validation is cheap and the miner already did the hard part.
If validation is the main cost, then we would expect incentives in the current situation to be similar to how they'd be post-SegWit.
In post-SegWit world, miners do have less incentive to send the signatures in a timely fashion, but I believe they still have "enough" incentive, for the following reason:
If I'm mining, and minerA sends me a block with no signatures, then minerB sends me a block with signatures, I'll switch over to mining on minerB's block. It doesn't cost me anything to switch, and I get more assurance that my block won't be orphaned. Even if minerA immediately sends me their signatures when they see that minerA sent their signatures out, I will still want to switch to minerB's block to punish minerA for being antisocial. Switching doesn't cost me anything, and it encourages a healthier, more valuable network. (Bitcoin is more valuable when blocks are publicly verifiable).
Further, what is the claimed incentive for the delaying miner to keep his signature data private for longer? Is the idea that this miner will eventually want to start mining invalid blocks, and he's priming the community to be vulnerable to this? What's the endgame here? That users will eventually just accept invalid blocks because it'd be too much of a hassle to unwind the chain?
Ask yourself: if you woke up tomorrow and you saw that the current longest chain had an invalid block from 8 hours ago, where a miner stole the funds of some random address, but the mining majority kept mining on this chain, would it change how you value tokens on that chain?
Another thing to note is that there's sort of an 'efficient market hypothesis' for miners wanting to validate blocks. When invalid blocks are so rare that they almost never happen, then the value of knowing whether the current chain is valid is low, because you can assume with high probability that it's valid. When miners start mining invalid blocks occasionally, the signature information becomes valuable since it lets you eliminate the risk of mining on an invalid chain. In that sense, this alleged problem is self-correcting.
