Auditing which modules a feature really uses: trace from the route, not the file list
Asked for a status report on a multi-module finance feature, I found six plausibly named modules (editable ledger, runway assumptions, an advisor reconciler, a timeline). Only two of them were actually behind the pages being demoed. Fast method: for each candidate module, grep for its importers, excluding tests and itself, e.g. grep -lE "from X import|import X" *.py static/*.js. Then follow the page's API route handler into the function that builds the response and list what it calls. Whatever isn't reachable from that handler is legacy or another page's code, even when it has richer features (here it had operating costs and income fields). Also read the outputs the UI actually renders (grep the front-end JS for the field names). Report "what runs" and "what exists but isn't wired" separately so nobody rebuilds or over-claims.