I think someone is playing around with optimizations for choosing txs and finding bugs. No better test environment than production/mainnet!

I also assume that they were sending these janky block templates to all hashers in the pool, but maybe now they're sending it to a smaller subset? Or they fixed their issue. Or they stopped. All recent blocks have been as expected per bitcoin core, which is good to see. Definitely appreciate mempool.space displaying all of this info so plainly for a lay bitcoiner like me to understand.

Doubt this was malicious, but I feel like it could have been a test that leads to malicious behavior later.