Kinship-aware null models (REML/PQL) are recomputed on every resume because the sparse Cholesky factor is not serializable in v1 of the null model cache (null_model_cache.rs line 10-12).
For large kinship analyses this wastes minutes on each restart. Serialize the factor or cache the derived quantities (residuals, variance components, projection matrices).
Kinship-aware null models (REML/PQL) are recomputed on every resume because the sparse Cholesky factor is not serializable in v1 of the null model cache (null_model_cache.rs line 10-12).
For large kinship analyses this wastes minutes on each restart. Serialize the factor or cache the derived quantities (residuals, variance components, projection matrices).