The solr_regex_replace processor runs on all stages: preprocess_index, preprocess_query, and postprocess_query. However, it seems the query stages have the side effect of corrupting searches when the regex is too broad. This happens even if the regex is restricted to only certain fields.
In my opinion, regex doesn't need to run on queries at all. Or, alternatively, the regex processor could be split into two, one for index time and one for query time.
Exact match should return results
Search returns no results
Active
4.0
Code
I think I agree on postprocess_query.
But I think it is wrong to only apply the regex at indexing time. Expect a few features like synonyms, terms should be treated identically during indexing and querying.
Unfortunately we don't have test coverage for the processor yet. Are you able to provide a failing test?