I tend to use lots of fundamental factors, set up elaborate pipelines, when writing my algorithms. I've been experiencing inconsistent behavior when I run a full backtest lately with my algorithms: Quantopian sometimes fails execution complaining about excessive memory usage or excessive time, sometimes not. With time especially, I've found that if I leave the browser window where I had started the full backtest, the backtest execution slows down. I don't have any hard evidence for this unfortunately, but I've found that when start a backtest, close the window, and then reopen the backtest, it seems to be running more slowly. Sometimes I find that the backtest appears to have stopped running altogether.
Does quantopian reduce the execution priority of a backtest that's not being directly observed? Is that taken into account with the timeouts that are set for the various function calls? In other words, does Quantopian measure wall clock time, or CPU clock time?