We shouldn’t spend more than 1 day on this.

Below are the groups I've found. I would suggest starting with the first group, possibly with a subtask for each of the feature files.

I see one subgroup about 'Termbox editing':

• label.feature (6 scenarios of which 4 are parameterized)
• description.feature (6 scenarios of which 4 are parameterized)
• aliases.feature (17 scenarios of which 4 are parameterized)
• header.feature (6 scenarios of which 2 are parameterized)
• empty_label_and_description.feature (8 scenarios of which none are parameterized)

Then there is a group of three tiny tests with a small single scenario each:

• non_existing_item.feature
• edit_tab.feature
• delete_item.feature

There are the two smoke tests which are somewhat cross-cutting:

• item_smoke.feature
• property_smoke.feature

There are the two about sitelinks:

• sitelinks.feature (16 scenarios)
• sitelinks_badge.feature ( 5 scenarios)

There is the long list of statement tests.

• statement.feature
• statements_geo.feature
• statements_item.feature
• statements_monolingual.feature
• statements_quantity.feature
• statements_rank.feature
• statements_snaktype.feature
• statements_string.feature
• statements_time.feature
• statements_url.feature

Lastly there are three that appear to be not part of any other group:

• reference.feature (12 scenarios of which 2 are parameterized)
• special_set_label.feature (7 scenarios)
• authority_control_gadget.feature (1 parameterized scenario)

The person working on these task should decide which scenarios have to actually be implemented and which are obsolete or redundant with already existing tests.

@Michael sounds good to me, thanks for the investigation. Would you please go ahead and create sub-tasks in T180228 with the groups you suggested? If that would take more than half an hour, please just move this to Done then and I will create those sub-tasks eventually.