Two issues so far:

1. The milestone itself doesn't get categorized as belonging to the milestone.

phab=# select date, id, title from task_history where source = 'ios' and milestone_title like '%Product Owner%';
        date         |   id   |                         title                         
---------------------+--------+-------------------------------------------------------
 2016-02-05 00:00:00 | 124303 | Migrate iOS app to send events to piwik.wikimedia.org
 2016-02-04 00:00:00 | 124303 | Migrate iOS app to send events to piwik.wikimedia.org
(2 rows)

2. Although two task.days do have that as the milestone, they don't get tagged as intended in the recat.

The second thing is not a bug. The recategorization goes in the sort_order from the recat file, and the first thing in the recat file is

1,True,App 5.0 is Released to Production,iOS-app-v5-production

phab=# select date, id, project, projectcolumn, title from task_history where source = 'ios' and milestone_title like '%Product Owner%';
        date         |   id   |        project        | projectcolumn |                         title                         
---------------------+--------+-----------------------+---------------+-------------------------------------------------------
 2016-02-05 00:00:00 | 124303 | iOS-app-v5-production |               | Migrate iOS app to send events to piwik.wikimedia.org
 2016-02-04 00:00:00 | 124303 | iOS-app-v5-production |               | Migrate iOS app to send events to piwik.wikimedia.org

So move that below Analytics (and probably below everything more specific) in the recat file.

The first issue is probably a corner case where Phlogiston doesn't consider the Milestone to be its own parent, and doesn't check the title of any task when recategorizing, and so misses all of the milestones themselves. I'll make a seperate bug because this is probably the right outcome (although the wrong way to get to it): if we counted the milestone task itself, then all Milestones would look like they had something in them and had a forecast, when really they don't and shouldn't.

Can you confirm that this is fixed? I'd like to be sure that it's different from T127497: iOS Milestones missing from Phlogiston. For the example case, I see that it has four sub-tasks in Phabricator, and it has a batching burnup in Phlogiston showing 4 tasks: http://phlogiston.wmflabs.org/ios_tranche10_burnup_count.png, so it seems fixed.

This original issue is fixed