101/102% matches are called by different names depending on the CAT tool. They are sometimes called Context Matches, Perfect Matches or ICE Matches.
When a segment is stored in our translation memory, we store not only the source and target text, we also store the content of the previous and following source segment. For example, this is what you could see in the TM.
<Previous Segment> I live in a small village.
<Source Segment> I have a small house. <Translated Segment> J'ai une petite maison.
<Following Segment> It is blue.
The translation is stored only for the segment that is being translated, but the 2 other segments are used to provide context. If that same segment was encountered again, and none of the 2 segments around it matched what is being stored, then we would have a 100% match because only the text matches. But if in the next document, one of the context sentences was also there, we would have a 101% match and if both were there, it would be a 102% match. Having the context sentences match what is stored in the TM helps increase the certainty that the translation is a perfect match for new segment.
In practice, 101/102% matches are often locked during pre-translation by project managers when the project is started because customers don't pay for these segments in most cases.