You really don't want to be measuring how well a dog recovers from being checked or those below average performances or measuring bad runs.
But if it's not checked in the first split you'd keep that data, if the check occurred after the first turn you don't need that 2nd split data or overall time.

Dogs that generally lead the race will run above avg performances, so if you expect a dog to lead, you might decide to only use data when it has raced 1,2,3 at the first turn.

A point to remember most dogs don't perform to a Normal distribution, and depending on the box position and racing style these play an important part of what shape distribution is required. This requires thought but it's logical