GET /api/rest/issues loads the data of each issue in the page one by one:
bug text, attachments, notes, relationships, monitors and history.
So the number of DB queries grows with the page size, and a page of
50 issues runs hundreds of queries.

Some queries are also repeated in the same request:

• tag_get_by_name() runs once for every tag line in the history,
  even for the same tag.
• When a project has no custom fields, an empty list is passed to
  custom_field_cache_array_rows(), which then loads all custom fields.

What needs to change:

• Load the data of all issues in the page at once, only for the
  requested fields.
• Cache tags found by tag_get_by_name() for the request.
• Skip loading custom fields when the project has none.
• Clear these caches whenever the data is written, so a write is
  seen by the next read.

The response must stay the same. The web UI would also get faster,
because the caches would be in the core APIs.