If I understand the current PoW requirement, it involves hashing (among other things) the previous block hash. Is there a (major) downside to replacing the previous block hash with the previous (or expected) account confirmation height?
This would still make the PoW unique enough that it couldn't be used multiple times.
The potential upside I see is as a spam deterrent. Right now a potential spammer is able to pre-compute blocks (and thus the required PoW) for a near infinite number of future blocks. Their future account actions are intentionally mapped out. A good actor, in almost all situations, does not have this ability. I don't know how many times I'll use my credit card today, twice? 5 times? 3 times real quick then 1 an hour later?
What I do know for certain is I will use my credit card again (Nano already allows for me to preemptively generate my new PoW). What I'm fairly certain of is I'll use my card 2 to 3 times per day and thus likely 15 to 20 times per week. I could never predict the exact details of all these transactions (I.E. in nano speak, I could never predict future block hashes) but I do know they would be the 375th to 395th transactions on my account (I.E. in nano speak I can easily predict my next confirmation heights)
If I could pre-compute the PoW for my next "X" transactions ("X" being of my or my wallet providers choosing") then I could either (without much consequence) choose to compute higher difficulty PoW or utilize slower PoW engines (mobile, CPU, etc), an option I'd never have for back to back transactions during a spam attack/saturated network.
I have to think a spammer would see the challenges of competing against a network full of intermittent users, who by being intermittent users have plenty of time to compute PoW. I made a lengthy Reddit post yesterday about spam mitigation and how IMO PoW calculation time is at the heart of spam prevention. The above seems to me to be a very Tai Chi type defense for most Nano users, utilizing their natural strength, idle TIME.