> if your contribution is not merged in, it should not count as "work done"

I highly disagree with this. Sometimes someone has to do the work to discover that isn't the work that should be done. As an example, last week, I got in a fight with the Go scheduler: https://github.com/php/frankenphp/pull/2016 -- in the end, we were able to find the one-liner that is a happy-medium. I didn't open that PR, but I did the work; if that makes sense.

In a program like this you can’t optimize for the assumption that every participant is acting in good faith and contributing good work even if it’s not accepted.

If a program incentivizes opening PRs even if they’re not accepted, the result will be a lot of maintainer spam from people opening useless PRs. This isn’t a personal hypothetical, it’s what we observe any time programs try to incentivize open source work. See the Hacktoberfest drama of years past where the promise of a T-shirt led to spam PRs across GitHub https://joel.net/how-one-guy-ruined-hacktoberfest2020-drama