The library VL database has PRU "% Return On Shareholders Equity" = 35% and net cash of $7.5Bn. Can't trust the data.

Indeed, can't trust it.
Goggling "pru roe" we see this: "During the past 13 years, Prudential Financial's highest ROE % was 15.71%. The lowest was -3.68%. And the median was 7.71%."

Various data sources say "Cash And Cash Equivalents" 14.6Bn (or 17.5Bn) and LTDebt 20.9Bn, thus a negative netcash. Both figures have been fairly stable in the last 5 quarters. And years.
See, for example, wsj.com - Balance sheet

I don't know how VL could be that far off. Maybe a transcription error?, either yours or theirs.
Hopefully one bad pick out of 40 (or 50) won't hurt too bad.

without a valuation component, and trust in the data source, I just won't be able to stick it out when it inevitably lags for a few years.

You mean like the Mechanical Investing screens such as YLDEARNYEAR? {ugh}

ROE, netcash, 5-year-sales-growth all seem to be valuation components.

At some point it becomes attractive to slow down on the frequent trading and move a chunk of money into something that doesn't trade very much, and just leave it sit.

FWIW (nothing!) the Jim-inspired 100-stock screen that I posted as "A new idea" has returned 2.33% vs SPY 0.88% since 7/3/23.

The (meaningless) 2 day return of my picks here is -0.40% vs SPY -1.42%.