I don't know all the answers, but I can address some of your questions.

Looking at this platform (PALnet) as an example... what’s to stop 💩posts from starting to use the tag, thus ‘polluting’?

Take a look at the top of the PAL richlist:

(source: https://steem-engine.rocks/tokens/PAL/richlist?sort_field=stake&sort_order=desc)

The highlighted accounts I know are controlled by their inner circle (and more could be... I'm not well-versed with MSP's leadership structure).
One of the advantages of a community launching its own token is that it can dedicate a meaningful pool of it to inside members and abuse-prevention accounts to enforce the community standards. With decentralized stake-based consensus, this is a necessary decision.

One of the problems on Steem is that Steemit did not fully anticipate the categories of abuse that would arise, and were not willing to commit enough resources to deal with them when they emerged. @steemcleaners has backing from Steemit, but other anti-abuse initiatives are fully community-funded and voluntary.

From a technical perspective.... can witnesses/devs run the numbers to see what will work?

There is a testnet and an API has been made available that allows witnesses and technically savvy community members to run scenarios and see how the reward pool would look under different scenarios. There are still some inherent limitations, though. We can test the rewards of existing behavior under new rules, but it doesn't anticipate how the rule changes will affect the existing behavior. That's a counter-factual that is not actually testable until the changes are made. (And that's what worries me about it. I'm hopeful about the impact of the changes, but hope is not a strategy. I don't think major changes should be made without a more solid basis than hope.)

people really do have different value-systems

Yes. I discoursed at length on 'what is quality' in @steembasicincome's second post on the EIP. Quality is in the eye of the beholder. That's why we have consensus-based rewards, though. Each stakeholder makes their own determination about whether the quality (or lack thereof) in a post merits them weighing in on its rewards. The vast majority of the time, a stakeholder will determine a post to be not worth voting, either up or down. That could be because they don't see it or because they just don't care. In some cases the stakeholder decides to vote for the post receiving a share of the reward pool. An in rare cases, the stakeholder may vote against the post receiving a share of the reward pool. It's a consensus mechanism, and a non-vote is perfectly acceptable - but in the case of a non-vote, you either concede the rewards decision to other stakeholders or ratify the existing consensus by opting not to change it.

I’m not a coder, so I don’t have an insight into that side of things.

I'm actually not a coder either... more like a systems analyst or design architect. I have a pretty good grasp of the logic and how all the pieces fit together, but I don't know any of the syntax.