So we know now why some test passes don't seem to get submitted to the wiki. What happens is this:

1. Scheduler submits jobs to openQA, gets a list of job IDs back
2. Scheduler sits there and waits for that list of job IDs to be completed
3. Some of the jobs fail / timeout for Mysterious openQA Reasons, openQA auto-restarts them
4. The restarted (cloned) jobs have different IDs
5. At some point all the original jobs are complete and the scheduler submits all the passes, but it does not 'see' any passes among the cloned (restarted) jobs, because it never got their IDs

There are two basic approaches to trying to fix this: we can try and spot when a job gets restarted and pick up the clone ID, or we can simply switch from the 'get the list of job IDs scheduled and wait for them to complete' approach to an approach where we just wait for all jobs for a given BUILD to be complete and them submit all the passes. The Python client actually already has a function that does exactly the job of 'wait for all jobs for a given BUILD to complete then return their IDs', so this would not be hard. It re-does the 'all jobs for BUILD' query each time it polls, so it picks up restarted jobs.

I'll look more into this tomorrow, if no-one else decides to pick it up sooner.