In version 7, duplicate result rows caused by strlst indexes (with indexvalues set to split, the default) are removed, so that results are more consistent with linear searches: a given row should be represented at most once in the results.
|Caveat: This means that many ORDER BY, relational etc. queries that use strlst indexes will have a lower total result count. The same rows will be returned, just without duplicates.|
This behavior change is rolled back if
<vxcp compatibilityversion> is
set to 6.