It turns out that the not scanner, (?!word3a|word3b) only works on lines, not files, even if you try ``(?-s)(?!word3a|word3b).

You’re misunderstanding what is going wrong. The expression you gave says, “Match null here if the following characters aren’t either word3a or word3b.” That matches at every position in a file except immediately before word3a or word3b. (Each line is listed once in the search results, but look at the number of hits.) What is needed is (?s)\A(?!.*(word3a|word3b)), which says, “Match null at the beginning of the document if the following characters don’t match any number of arbitrary characters followed by either word3a or word3b.”

When I tested the whole expression I got some erratic results when I had it match null, which is why I set it to match the string from the beginning up to the word word1.

looking for regex which can find files which satisfy this criteria (contains word1) & (contains word2a or word2b or word2c) & (does not contain word3a and word3b). In above example it should find only File 3 and File 4

Try:

(?s)\A(?=.*(word2a|word2b|word2c))(?!.*(word3a|word3b)).*word1

Edit to add:

(?s)\A(?=.*?\b(word2a|word2b|word2c)\b)(?!.*?\b(word3a|word3b)\b).*?\bword1\b

is probably better. The \b assertions require word boundaries — so that reword2a or word13 won’t count. Changing .* to .*? won’t change the results, but it might make the expressions more efficient and less likely to run into “expression too complex” failures.

My passes were:

1. Scan for files containing word1. Either make a list of these. It turns out all of your files contain word1.
2. Scan the list of files from step 1 for (word2a|word2b|word2c) - We are down to a second list that has files 1, 2, 3, and 4.
3. It turns out that the not scanner, (?!word3a|word3b) only works on lines, not files, even if you try ``(?-s)(?!word3a|word3b). Thus I would scan list 2 for (word3a|word3b)` and exclude those. This excludes files 1, 2, and also 5 and 6 but those are not in list 2. The remaining files are 3 and 4.