While working on T328913 I have identified the following issues with content translation events.

Time period reviewed: 90 days preceding 15 December 2023 [274525 events across 28260 unique sessions]

content_translation_session_position

• Issues
  1. Within a session, various event types have same session position, even though they occurred at different times. For example in one of the sessions, dashboard_open, dashboard_translation_start, and editor_segment_add, all have session position as 1.
     1. ~54% of the sessions have this issue.
  2. Within a session, events with same event type that occurred at the occurred at different times, have same session position.
     1. ~40% of the sessions have this issue.
  3. content_translation_session_position = 0 has event types other than dashboard_open. I am not sure if it is possible to open the editor or translation start page without opening the dashboard.
     1. ~1% of the sessions have this issue.
• Expected behaviour
  • Ideally, within a session (content_translation_session_id), content_translation_session_position should be unique, even for same event_type. This has to be an incremental integer to understand the sequence of events.
  • As the first event to occur has to be dashboard_open (if that's the case), other event types should not be logged for content_translation_session_position = 0

user_global_edit_count_bucket

• Issue
  1. There are instances where within a same session, different values were recorded for user_global_edit_count_bucket. For example, in one of the sessions, the edit count bucket had both 0 and 1000+ edits values. I am not sure if this will be true for all the cases, but the ones I have observed, the bucket is being set to 1000+ edits, when user_global_edit_count is null.
     1. ~7% of the sessions have this issue.
• Expected behaviour
  1. If this being caused due to user_global_edit_count being null, user_global_edit_count_bucket should be null as well.

number of events produced in an hour

• While it is hard to say how many events can practically be produced within an hour, there are a few sessions with very high number of events produced in one hour.
  • 82 sessions have more than 100 events in an hour
  • 16 sessions have more than 250 events in an hour
  • 6 sessions have more than 500 events in an hour
  • 1 session has more than 1000 events in an hour

expiration of content_translation_session_id

• As per documentation, content_translation_session_id should expire after 15 minutes if not events are produced in that stream. This doesn't seem to be happening. I am not sure how many sessions have this issue, but I did a test, and even though the events were triggered 15 minutes apart they have the same content_translation_session_id.