This matters more for backtesting than live execution, and it's where systematic work quietly breaks. When you stitch ESM26 into ESU26 to build one continuous chart, the two contracts don't trade at the same price, so you get a fake gap at the splice from cost of carry, not from any real move.
Back-adjusting removes the gap but shifts every historical price, so your absolute levels stop matching what actually printed. Ratio-adjusting keeps the shape but distorts the distances. Neither is free. Anything that reads off fixed price levels instead of relative structure is exactly where a backtest diverges from what you'd have actually filled, and the roll is usually the culprit.
Liquidity migration is the other piece. Volume drains out of the front month into the new one across the roll window, so the levels worth watching shift to ESU26 before the calendar says the contract officially changed.