Problem

When MT used for pre-translation of segments with repetitions master translation may have different set of tags compared to expected for repetition: for example segment with content protection tag was pre-translated by MT resource without CP tag, then master translation is still returned as best match for translation.

This results in inconsistent translation.

Solution

In case if repetition handler fails to apply master translation for repetition translate repetition with additional call to MT.

Test

1. Enable integer CP rule for language pair
2. set "runtimeOptions.LanguageResources.deepl.tagDialect" to "none"
3. Create task with 2 segments that have different integers
4. Pre-translate the task with DeepL

Expected in master:

Repetition has same translation as master with wrong integers

Expected in fix:

Repetition has own translation as master with correct integers