The test cases above are good in general, as these test a small aspect.
Improvement:
These are not different test cases, but iterations of the same test. This calls for data driven testing, see:
http://doc.qt.io/qt-5/qttestlib-tutorial2-example.html

I think this test should be written in a different way:

• check if the map has the correct size
• check if each entry has the correct docId
• check the terms:

for (const auto& feat : { "no", "ot", "te", "es"} ) {
    QCOMPARE(exported[feat].size(), 1); // one document
    QCOMPARE(exported[feat][0].wid, 0); // first word
    QCOMPARE(exported[feat][0].len, 5); // strlen("notes")
}
...
QCOMPARE(exported["md"][0].wid, 3);
QCOMPARE(exported["md"][0].len, 2);

You should have a separate test case here, or even 2 - you should test if the merging works as expected:

• for same feats from different documents ("notes")
• for same feat from one documents ("not_es_", "wedn_es_day", i.e. "es")

If you create a list of the files, you can use a loop for the insertion here.

QList<QStringList> files = {
    {"notes", "april8", "2018", "md"},
    {"notes", "wednesday", "04092018", "md"},
...
    {"LMC2200"}
}
...
for (size_t i = 0; i < files.size(); i++) {
  db.unite(FuzzySearch::features(i, files[i]));
}

This function itself calls for a unit test - for a given feature, return the matching documents
Also, either capture the list, or return it (then, by value, not by reference), but don't do both. I strongly prefer return by value here.

If you create a QSet<QStringlist> foundFiles = files[result[0]];, you can compare it with QSet<QStringList>({{"notes", "wednesday", "04092018", "md"}});
This way, it becomes obvious what you expect as output here.
To make it more obvious what the QStringList refers to, you can use using FileNameTerms = QStringList;