Yeah, I'd always take the points you mentioned into account when reviewing a solution. Along with saving time I did find reviewing Codility submissions quite revealing. I've seen cases where the candidate couldn't get code to even compile, another where they didn't know Java includes map data structures existed (asked them in the follow up interview) and another where the implementation for an easy task was super slow when the nearly fastest approach should have been obvious (again, asked about in interview).

I'd say it's better for identifying really unsuitable candidates than it is for identifying who are the best ones.