Corporate actions are data quality events, not accounting footnotes

Following the adjusted-price thread, the broader version: any corporate action is a moment where your historical series and reality diverge unless something reconciles them. Splits are the obvious one and the easiest to catch, because the price jumps visibly. The dangerous ones are quieter. A dividend shifts the price by a small amount that looks like an ordinary move. A ticker change breaks the link between old and new history entirely, and depending on your data source you either lose the past or silently inherit a different company's. A merger ends a series mid-stream. The reason to care: every one of these creates a fake signal. A dividend-driven price drop looks like a gap down. A ticker reuse gives you a "history" belonging to something else. And unlike a bug in your code, none of it produces an error. The habit that helps is boring. Keep a log of corporate actions affecting your universe, and when a backtest result looks unusual, check that log before you check your logic.