Multiple translations and duplicates

I am going to explain this problem and also describe how it is handled in three main CAT tools. First, some definitions, as these terms are used differently, and it is easy to misunderstand the issues:
Multiple translations: Same source segment translated differently in TM/MDB
Duplicate: Same source segment in the document
File duplicate: Same source segment in the same file
Project duplicate: Same source segment in the same project
General duplicate: Same source segment in files from different projects
Some historical references
If I remember things correctly (which I sometimes do not), it was like this:
The first versions of TRADOS did not produce multiple translations if you edited a segment, for which there was a 100 % hit in the memory. As some translators wanted this to happen, this was changed in one of the updates. There were a lot of complaints on this change, as many translators (including me) wants to avoid multiple translations in most cases. If you work with highly repetitive material from the same sources year after year, you would appreciate not to have a pollution of the memory by segments that are edited or corrected.
Then Trados made the possibility to choose approach, by setting Multiple translations: No and Options: Do not create new translation units if only text fields differ. I then reorganized my databases and removed index fields, and only used a text field for the project number, then it was possible to stop the memory pollution. This is a Workbench setting, so it works both in TagEditor and when working with Word + Workbench.
The upgrade to Studio showed that SDL/Trados almost completely had forgotten about this debate, and the Atril people never had it.
Studio:
Creation of multiple translations
If you come to a line in the translation that has a different translation in memory, there is a little yellow triangle showing below the translation results window. Clicking on that triangle gives you a warning message. But if you do not observe that, the default is that a new memory entry is produced.
Checking multiple translations
If you already made the mistake, nothing shows that the same source is translated differently. The verify function does not check this. There is a possibility to filter on Duplicate segments, and thus eliminate multiple translations for file duplicates, but not for project or general duplicates.

Atril DVX
Creation of multiple translations
There is a protection against file duplicates, and also a claim that project duplicates are protected. But if you add a new file in an old DVX project, then that protection is overridden.
If you come to a line in the translation that has a different translation in memory, there is no warning message and a new memory entry is produced.
Checking multiple translations
If you already made the mistake, nothing shows that the same source is translated differently. The QA function does not check this. There is a possibility to filter on Duplicate segments and also for duplicate segments with different translations, and thus eliminate multiple translations for file duplicates, and probably for project duplicates if you run all files in the project in the editor, but not for general duplicates.
Old Trados
Creation of multiple translations
It is possible to choose setting for this, and multiple translations can easily be avoided as the default is that memory is changed.
Checking multiple translations
Translate until fuzzy resolves all kinds of differences between memory and document. This kind of function that opens unit by until, compares document to memory and reports differences unfortunately is missing in both Studio and DVX
Autopropagate
The remedy for keeping the translation consistent in Studio and DVX seems to be a function called Autopropagate. I have tried it in both program, but it is complicated to set up and it is limited to file duplicates (in DVX also project duplicates) but cannot handle general duplicates. And in the present version of Studio the Autopropagate is useless because of the numbers/codes corruption. If you have Autopropagate turned on, and there is codes like 2.3.4, they are changed in all segments to 2,3.4 (Swedish language). This will probably be fixed in next service pack.
There are practical problems with the Autopropagate. If you choose prompting for changes, and have 20 occurrences of same headers, the prompt will turn up 20X20 times and you will have 400 extra clicks which is not a productivity enhancement. If you have automatic update, and then there actually *is* a few units that must have different translations in the same file, then you get into trouble.
My translation model
The way I mainly work with translation may be somewhat old-fashioned, but I actually prefer to work without pretranslation and propagation, I want to start with the first line, then take the next, and get the hits from memory, edit all hits, when needed, and if I edit an 100% hit I will have the edition in memory.
After translation I would like to have a consistency check to see that all duplicates are translated the same and that all sentences are consistent with the memory.
If I make terminology changes with search and replace,, I need a simple method of browsing through the document that lets me edit and resave changed TU:S
This very simple and straight forward MO was possible in Trados, but unfortunately is not in Studio or DVX. The main problems are the multiple translation production, and the lack of methods to compare document to memory.
Both function seems to be easy to implement at least as an option, as there is possibilites in both programs for this, which are not utilized.