It seems to me that sendtemplate is just transaction relay with extra steps. I don’t see the point, we should just focus on transaction relay instead, by increasing the size of the orphan pool, for instance.
A template (without using weak blocks) is costless to generate so is just an “extra step”. There’s also no way to know whether the sender is actually a miner and trying to mine the block. If I want to fill your orphan pool with junk that wouldn’t normally be relayed, I’ll just send all my peers templates with the junk txs. I think we need to think harder about relay policies against DoS, and instead using a (very) large orphan pool and blockreconstructionextratxn.
I posted [some comments on the PR](bip-ajtowns-sendtemplate by ajtowns · Pull Request #7 · ajtowns/bips · GitHub) that I doubt anyone read. and since learned about blockreconstructionextratxn…