Is the following scenario valid?
1. Some unhonest segwit mining pool takes top-1000 segwit utxo and mines a block at height N with a transaction which transfers all funds to his p2pkh address
2. This block does not have segwit data portion, but it can be broadcasted to all non-segwit nodes on the network
3. All other pools have a dilema - wait the segwit data associated with this block or start mining block N+1 on the top of N
4. What if miners will use SPV-mining on the top of this block? They will create blocks at heights N+1, N+2... etc without checking the segwit-validity of block N
No different than the situation today with "spend all the coins in the first 10000 blocks, without knowing the private key; and hope that miners extend the chain without having or checking the block. The segwit data is not sent separately.
In either case the corrupt chain would be simply reorged out after miners hit their limit of mining without data or otherwise some process realizes they are mining a chain the network is rejecting. Non-validating clients would be exposed to the reorganization, ones that are validating would not be.