Underleveling and barriers to promotion are rife at Google.

I wonder what doors get opened if someone successfully sues for being underleveled.

Under leveling compared to what?

It's a common problem in many large companies that have significant barriers to handing out promotions, like google's need to work on promotion packets, and passing past review processes. Often you will find that output and level have little to do with each other, including within a team.

Where I am working, my lowest level teammate is clearly among the most competent by most standards. The one person I'd delegate a task to and expect it to not just get marked as done, but actually get done. I have coworkers in the team multiple levels higher that might produce basically nothing in any 2 week sprint, and it's not because they are spending time in meetings and coordination: They just have minimal productivity. But getting rid of the worst is a major hassle, and the middling ones aren't worth getting a PIP. Other teams are in better positions to ask for uplevels, because they do more important things. Therefore, the only sensible approaches are to either take it very easy, lowering the performance curve, or changing teams/jobs to places where one can be upleveled properly.

It's not as if making it easy to fire and promote is a panacea: I've seen managers laying off their best workers because they were politically threatening, and teams getting little done that had everyone marked as a staff engineers. Ultimately performance is hard, and being bad at it, like keeping people underleveled for years, makes the organization degrade. Top paying companies like Google aren't immune.

Of course they have those barriers.

The incentive for managers is to promote.

Perhaps if managers had budgets of dollars instead of budgets of headcount then the imbalance wouldn't be as large, but AFAIK companies do not do that except at very high levels.

Having a dollar budget would make the imbalance worse unless ideally they don't get away with moral hazard. The issue is usually having managers taking all the benefits and none of the risk, and that's bad.

Fellow employees who do similar work at similar levels of quality.

Qualitative measures are not quantifiable. Good luck in court.

It's not much of a caricature to say American courts discovered numbers circa the 1940s (cf Learned Hand's introduction of the concept in antritrust and negligence cases); before that they did just fine.