# Reproduction and release limitations

- One article example was reproduced; the entire Walker study was not rerun.
- Authentic archived inputs were already harmonized. Variant discovery, LD
  clumping and source-database queries were not reconstructed.
- The archive omits an embedded exposure unit. The current catalogue reports SD;
  this wording is retained rather than retrofitted into the source bytes.
- The original RNG seed and R 3.5 environment are unavailable. Fixed-seed current
  and historical-package compatibility runs cannot recover an unknown seed.
- Bootstrap median/mode uncertainty is stochastic and was assessed against the
  prespecified 20-seed envelope; it was not forced to a displayed value.
- Heterogeneity, Egger-intercept, leave-one-out and funnel displays diagnose
  patterns within their assumptions and power; none proves instrument validity.
- Technical replay does not establish biological causality, clinical efficacy,
  safe intervention, institutional approval or publication readiness.
- Advanced correlated, multivariable, one-sample and colocalization workflows are
  outside the qualified core.
- Original article/figure files, author code, authentic association tables and
  historical software archives are not redistributed in the public package.
- The open-preview license decision is operational, not legal advice. Counsel and
  institutional ownership review have not occurred.
- Price, support, updates, external marketplace terms and buyer-delivery behavior
  are unresolved; no checkout or paid delivery is enabled.