The command was modified to use the getBuildLogs call in #3232, but if the build has a task id, then it will instead just download task outputs. Since task outputs tend to get cleaned up fairly quickly, this preference seems backwards.

It's a little unclear what the best approach should be. There are cases where getBuildLogs could fail to logs, but listTaskOutput might, e.g. for a recently failed build.

I suggest that the command use getBuildLogs for builds that are in the complete state whether there is a task id or not, and only fall back to the looking up the task output if the build is in a different state.