Well, my exact use case, which I imagine I'm not alone in, is indexing a tab-delimited file with a single-line header containing the column names. So indexing by column index would be adequate, but it would present two very minor problems: 1) I'd have to figure out a way to exclude the header row from output (or alternately, to always output it), and 2) it would be slightly more convenient to refer to the columns to be indexed by name rather than position, as some of these have thousands of columns.
Can you have multiple indexes on the same GZIP file? And if so, does it create multiple .zindex files or re-use the same one?
I don't know if SQLite supports custom data retrieval backends like Postgres, but what would be REALLY interesting is to write one such that the indexing information is in SQLite, and you can use SQL to query row data directly from the GZIP index (I know this is beyond the scope of your project, I'm just thinking out loud now).