Simplify existence checks for input datasets.

Logic branches for unresolved dataset refs in quanta and the old
version of pipeline mocking are no longer relevant.

python/lsst/ctrl/mpexec/singleQuantumExecutor.py

@@ -401,39 +401,15 @@ def updatedQuantumInputs(
             newRefsForDatasetType = updatedInputs[key]
             stored = limited_butler.stored_many(refsForDatasetType)
             for ref in refsForDatasetType:
-                # Inputs may already be resolved even if they do not exist, but
-                # we have to re-resolve them because IDs are ignored on output.
-                # Check datastore for existence first to cover calibration
-                # dataset types, as they would need a timespan for findDataset.
-                resolvedRef: DatasetRef | None
                 if stored[ref]:
-                    resolvedRef = ref
-                elif self.butler is not None:
-                    # This branch is for mock execution only which does not
-                    # generate actual outputs, only adds datasets to registry.
-                    resolvedRef = self.butler.registry.findDataset(ref.datasetType, ref.dataId)
-                    if resolvedRef is None:
-                        _LOG.info("No dataset found for %s", ref)
-                        continue
-                    else:
-                        _LOG.debug("Updated dataset ID for %s", ref)
+                    newRefsForDatasetType.append(ref)
                 else:
-                    # QBB with missing intermediate
-                    _LOG.info("No dataset found for %s", ref)
+                    # This should only happen if a predicted intermediate was
+                    # not actually produced upstream, but
+                    # datastore misconfigurations can unfortunately also land
+                    # us here.
+                    _LOG.info("No dataset artifact found for %s", ref)
                     continue
-
-                if (ref_stored := stored.get(resolvedRef)) or (
-                    ref_stored is None and limited_butler.stored(resolvedRef)
-                ):
-                    # We need to ask datastore if the dataset actually exists
-                    # because the Registry of a local "execution butler"
-                    # cannot know this (because we prepopulate it with all of
-                    # the datasets that might be created). Either we have
-                    # already checked and know the answer, or the resolved
-                    # ref differed from the original and we have to ask
-                    # explicitly for that.
-                    newRefsForDatasetType.append(resolvedRef)
-
             if len(newRefsForDatasetType) != len(refsForDatasetType):
                 anyChanges = True
         # If we removed any input datasets, let the task check if it has enough