Okay, let me first describe the scenario:
I assume the Undo-Command would not have worked here, would it? I noted this error only after doing a bunch of corrections on some selected segments.
My solution was to "Remove target segments same as target" and to pretranslate based on the TM.
This meant: the file was only 50 % translated, a whole bunch of 95 % matches (formerly translated) or 100% with different source segments (due to spaces etc.) that did not pretranslate automatically. A whole hour of extra work (thank God the file did not contain many tags). Some segments or only present as fragments inside the TM (any chance to explain this?), and a lot of numbers were puzzled.
What workflow would you have suggested in this situation?
To have a tool that does not automatically insert 100 % matches entered only some minutes ago makes me feel strange, but I assume there are some levers to move to change this behavior – as far as I know I already did, but it would be nice to have it explained again – missing the wiki.
It would be nice to have some more prompts for this kind of operation to avoid this kind of situations, in the kind of "This will affect all segments. Continue anyway?" with an optional "Do not show this prompt again".
'Processing tags' is always checked. This cannot be the problem.
Have a nice time there, hope it does not rain too much.