recording-based backtests: how many days before i trust it?

read the guide on backtesting against your own recording instead of resampled bars. makes sense to me in principle, but i have questions before i rebuild my whole routine. my situation: - i trade premarket into the first 30-40 min, mostly rotation off a gap scanner - universe changes every single morning, which is exactly the thing the guide says resampled bars destroy - so far i've been backtesting on 1m bars from a separate data pull and the fills always looked nicer than reality what i'm unclear on: 1. recordings only start once you're subscribed to a symbol. for me the symbol i care about often isn't in my list until the scanner pulls it in at 6am. does the recording capture it from that moment forward, and is a partial morning still usable for a backtest, or is it garbage in? 2. how many recorded days is a sane minimum before the overfitting check on the report is even meaningful? i can imagine getting a clean equity curve off two weeks and fooling myself completely. 3. "prepare engine" - do i need to redo that every time i add new recorded days, or does it build on what's there? not asking anyone to hand me a strategy. just want to know if my morning workflow breaks the assumptions the guide is built on.