Thoughts to date

Yes, a "safe enough to eat" is exactly what's meant. Not necessarily "Good Article" but meets a set baseline for quality.

1. What it gets us: Much easier to drive or promote basic improvements, and also educates the new users who write them. Hence likely to be an area we can make good inroads, automatically, on a large scale, engaging the wider public, related to what critics most notice. Each a distinct "positive".
2. Example criteria (random ideas): Decent feedback from at least 30 readers... sufficient word:citation ratio in each section... fewer than X tags per 1000 words for major issues (cite needed, sourcing, npov, etc)... no major article quality tags in the last X days... Y thousand page views (to ensure sufficient eyeballs to have a good chance errors were noticed). Borrow ideas from Good Article Criteria and figure which are essential baselines, which we need but not as strictly, and which we can somehow approximate by automation.
3. What we do with sub-baseline content: We set up automated systems, "Help fix this!" buttons when someone views an article, feeds for individual substandard issues, "Fix a random issue" button, everything we darn can, and drive like hell that EVERYONE can help fix basics in articles, readers, people who've never used Wikipedia before, ANYONE. "You can look up a citation, here's how!" ... "You can check a fact, or if a statement/section is fairly tagged, here's how!" ... "You just wrote an article, and I noticed some improvements that will help it stay on Wikipedia. Here are the top 2 items!" ... "This article has requests for help that match your filters, do you want to read them?" ... "This article is only rated at 2.1 for quality. Click to see if you can help Wikipedia with any of these issues" ... We push like hell for it, using automated methods, to get this kind of work automatic. That's what we do.
4. Inherent fails will usually meet a brick wall as usual.