An interesting proposal! Appreciate you referring to my post about Blob Aggregation. It was designed as an auxiliary solution (without changing the protocol core) to solve the blob underutilization problems today with the introduction of additional parties. We also thought about implementing the final solution at the protocol level, but we didn’t explore it as deeply or technically as you did, @1xiaoxiao11111111. 
Based on my understanding the proposal is viable (though requires a hard fork) and it makes sense. With introduction of dynamic blob sizes I believe it would make sense to remove the max_blob_number per block constraint and rather adopt a different metric like max_blob_size to constraint the amount of data that can be written on chain per block (like gas limit for blocks), rather than limiting the number of blobs that can post data.
I believe padding the last blob field is a clever solution as it keeps the expected blob structure for the different DAS approaches, while only truncating the redundant zeroes.