Read the original workflow configuration, pinned selection, and candidate checks saved during File-catalog resolution. No remote queries or candidate recalculation are performed. Missing records in older runs remain explicitly unavailable.
shift_selection(x)A list with state, run_id, requested, resolved, attempts,
cases, and diagnostics. Batches return a named list per child.
Each attempt identifies its index node and outcome. Its checks retain
candidate tables before and after service resolution, query IDs, required
variables and explicit constraints. Future candidate rows carry a reason.
The scope is the File-level candidates evaluated for requested models.
It does not include every model in ESGF, candidates excluded by upstream
Dataset discovery or server filters, or a history of arbitrary user filters.
eligible_alternative_not_selected describes execution selection, not a
scientific ranking. future_coverage_incomplete and
no_complete_matching_reference describe coverage or input pairing.
user_member_constraint and user_grid_constraint identify explicit limits
when those alternatives occur in the evaluated table. An empty check list
means failure occurred before candidates could be evaluated. An attempt may
fail after selecting candidates, for example during service access.
Plans return state = "planned". Finished runs without captured candidates
return state = "not_recorded", retaining any saved configuration and final
selection. Captured attempts return state = "recorded", which does not
imply successful weather generation. Inspect the attempt outcome and cases.