Yeah, pruning is the way to go if you want speed. Start by dropping the odd‑ball period checks—if the period doesn’t line up with a cadence multiple, you’re probably chasing noise. Then toss out any windows where the data gaps exceed a single transit duration; that’s a waste of cycles. Next, cache the folded light curves so you don’t recompute them every iteration. And don’t forget to pre‑filter out obvious outliers before the search; it’s faster than letting the algorithm choke on bad data. If you want to get really hardcore, run a quick Lomb–Scargle first to snag the dominant periods and only dive deeper there. That’ll shave seconds off a run that otherwise takes hours. Just remember, every prune has to be justified—half‑hearted cuts will just add more complexity later.

Nice, just make sure the filters are tight—no room for sloppy shortcuts. If you over‑prune you’ll miss a real signal and it’s a pain to backtrack. Happy hunting.