Your query using IntersectScans, You should avoid intersectScan (How can an Intersect Scan return more items than MIN(inputs from respective indices). It's returning the MAX) by dropping unnecessary index or specify USE INDEX clause
mdata AS o USE INDEX (ix_subset_version_model_fcstLen_fcstValidEpoch_DD_model)
JOIN mdata AS m USE INDEX (ix_subset_version_model_fcstLen_fcstValidEpoch_DD_model)
Second JOIN produced documents, it needs to walk thorough two arrays and construct new arrays,
Later it throwing away Filter time 95% of documents are eliminated might be due to length of the array check.
That why i suggested in other post use Array index.