problem

Using pre-translation with internal fuzzies enabled may lead to missing pre-translations on repeated segments. This happens only, if the real match found from the TM and the generated fuzzy match has both an equal match-rate >= 100.

reason

The reason is an implementation which groups the match results with the same target and match rate coming from the same language resource.

This code is implemented in the opentm2 connector, and does not know therefore anything about internal fuzzies etc. Also this code is only relevant for the fuzzy search from UI, not for pre-translation, since in pre-translation always one, the best, match is used (and fuzzies are already excluded there).

The code (mainly in function editor_Services_OpenTM2_Connector::getResultListGrouped()) belongs to business logik on sorting/grouping the results for the match grid in the UI, so it should be moved to a code level (into the controller, or even into the UI itself). Currently there is a config 
runtimeOptions.LanguageResources.opentm2.showMultiple100PercentMatches which could be omitted if implementing it directly in the UI (by just displaying there a checkbox or so).

solution

The correct solution is pointed out in the above reason.

A workaround currently applied is to sort the internal fuzzy matches down in the currently used grouping function.

The final solution must be discussed with marcmittag