How important is the quality of historical market data?

Backtests are only as reliable as the data and assumptions behind them. Missing candles, inaccurate prices, and unrealistic execution assumptions can distort the results. What checks would you perform before trusting the output of a historical trading test?