Q500US like screening in live/backtest mode

[kencheung0804]

I have a postgres database containing several thousands stocks' daily ohlc and fundamental data. My goal is to perform something like the old quantopian Q500US screening before other strategies. This leads to the following three questions I'm gonna ask:

First, especially in live mode, stocks change/ get delisted/ added rapidly. I know as long as I add all stocks into datafeed (assuming my ram is enough), delisted stocks are easier to handle. However, how can I add new stocks into datafeeds after run()?

This leads to my next question. I don't want to have survivor bias. However, is there any way I can perform the screening without adding all several thousands stock data, not matter delisted or not, into the datafeeds? I'm always expanding my database. And I'm afraid I'll end up lacking Ram.

Third, this is a general question. Is there any way I can mimic zipline's pipeline in backtrader live/backtest mode?

Thank you.