Task completion would also be abused, because it's too vague. It simply shifts the burden to the one formulating the task (e.g. preventing holes in the specification like missing performance or hardware requirements).
Exaggerated example: if the "task" is to automatically deliver a report containing certain data and formatted in a specified way, the easy way might be an implementation that stalls the DB server for hours with deeply nested FULL OUTER JOINs on non-index fields.
The task would be completed quickly and arguably correctly, since neither runtime nor memory requirements were explicitly specified...
But you said it yourself, any system is going to be abused...
Task completion would also be abused, because it's too vague. It simply shifts the burden to the one formulating the task (e.g. preventing holes in the specification like missing performance or hardware requirements).
Exaggerated example: if the "task" is to automatically deliver a report containing certain data and formatted in a specified way, the easy way might be an implementation that stalls the DB server for hours with deeply nested FULL OUTER JOINs on non-index fields.
The task would be completed quickly and arguably correctly, since neither runtime nor memory requirements were explicitly specified...
But you said it yourself, any system is going to be abused...