Problem

While converting JQL query to Lucene query Jira tries to guess projects to which the results will belong. This enables Jira to perform some optimisations. For example:

• Optimised security qeury
• version functions are expanded to versions from only related projects.

Unfortunately, those guessed projects are not passed into subqueries.
The problem reveals itself on Kanban board with default Kanban sub-filter:

fixVersion in unreleasedVersions() OR fixVersion is EMPTY

Test case

• Create 2 Kanban projects
• Create 33k unreleased versions for each project (66k in total)
• Try opening Kanban board

Expected result

Everything works. Lucene query contains 33k versions from current project. (optimization worked and sub-filter receives the project scope)

Actual result

Jira tries to create Lucene query with 66k versions from the whole instance. (Optimization failed)
It also fails due to BooleanQuery limit.

Workaround

Removing 'OR fixVersion is EMPTY' fixes the optimisation.