We have observed weird behavior in the Teradata environment , Basically some of the SQLs suddenly starts running slow and when checked and compared to the previous explain plan for which it ran with optimal efficiency , there was deviation in the number of steps and even the plan was slightly different.
Did anyone ever faced similar situation?
Usualy you find this kind of behaviour in case of missing / stale stats.
And the cost based optimizer might decide to chose a different plan in case the demographics change. To an extreme one single additional row might change a plan...