The stacked updates issue as in the link mentioned https://issues.apache.org/jira/browse/LUCENE-4258 handles FieldUpdates only for "new incoming values".
In our case, all fields that are updated are, by default StoredFields. Currently StackedTermsEnum looks too costly on computing Term Stats. Is it possible for us to read the stored-fields value during BufferedDeletesStream.applyDeletes() call [during flush/commit etc...], so that we can "adjust" TermStats etc... correctly during index write itself? We have our own codec holding TermStats, where such "adjustments" are possible [Simple Key-Value pair] One major issue is tokenization, which will delay NRT re-opens for a long time. Any suggestion is really helpful -- Ravi