Not all spurious correlation is random
People tend to talk about spurious correlation that drives from pure luck/randomness. Two unrelated timeseries end up correlated out of pure luck. The funny part is that this actually is a very low probability event for any two random timeseries that you pick, if they were both stationary. Sure the correlation might be somewhat positive or negative, but a 90%+ R^2 is actually quite unlikely for two completely random, stationary timeseries.
The thing that is actually more at fault a lot of the time is non-stationarity. Non-stationary timeseries have a tendency to be correlated between they are often trending in some direction, which tends to pull correlation away from zero mathematically. Non-stationarity is extremely common (Bet on non-stationarity).