Every claim below comes from the corpus's own verification report. For researchers, these are open questions. For funders, each one is a candidate for support: a place where a modest grant buys measurement that currently does not exist.
The pesticide-use record stops where the questions start: EPA's national series ended in 2012, leaving model-based USGS county estimates as the only continuation. Agricultural energy analysis rests on 2001–2012 data with no current federal replacement. Federal aquaculture data is split between USDA ($1.9B farm-gate, 2023) and NOAA (663M lb, $1.7B, 2022) on non-reconcilable definitions. Cotton and rice lack dedicated annual highlights. And a searching pass found no verifiable production-side research from Kresge, RWJF, Ford, or Packard — their food portfolios concentrate in health and equity, leaving production philanthropy to a much shorter bench. One catalogued Forest Service heirs'-property paper carries internally inconsistent aggregate dollar figures; the entry says so.
State-agency evidence is the thinnest tier in the corpus: sub-federal knowledge leans almost entirely on university centers (MSU CRFS, Johns Hopkins CLF, Harvard FLPC, Leopold). Trade policy is covered only through USDA outlook work — no primary USTR documents. The SNAP evaluation literature is represented by ERS reports, not the underlying Mathematica studies. No verified McKnight or Ford food-systems reports could be located. Thirteen entries are honest placeholders: directory and data-portal pages carrying "n.a." rather than borrowed numbers.
The case-study record has named, specific holes: Shepherd's Grain and La Montañita lack standalone verified write-ups; Minneapolis Homegrown and current Philadelphia food-policy reports were not obtainable; Wallace Center hub publications and Walton Family Foundation grantee evaluations remain unfetched; Fifth Season Cooperative's post-2010 outcomes are undocumented in any source that could be verified. Several widely cited program portals (Growing Food Connections, Agriculture of the Middle, WKKF food systems) turn out to have no citable data on their landing pages at all — the evidence, if it exists, is one unfetched PDF deeper.
Three grant-shaped conclusions. A state-agency sweep of ten high-agriculture states would strengthen the corpus's weakest tier for the cost of a single conference. A foundation-grantee retrieval project — pulling the actual evaluations behind Kresge, RWJF, Ford, Packard, McKnight and Walton program pages — would either surface a hidden literature or prove it was never written; both findings matter. An energy-model refresh would replace the single most-cited stale number in food-systems work (the 10:1 ratio) with a current one. The commons will catalogue, grade, and host whatever such work produces.
As set out in the verification report: repair URL rot with a browser-realistic re-fetch of all 254 sources; pull the PDFs behind landing-page-only entries; run the foundation-grantee deep dive; deepen the state-agency tier; update or formally deprecate the 2012 energy basis; build the six thematic dossiers; and write the long-form case studies for the successes and failures Domain C already indexes.
The complete QA record — URL sample results, grade audit, model-by-model honesty log — is in the verification report.