Why
Track coverage deferred from #143 so that the fix stays focused on Drafts swipe deletion and removing the exactly owned local IMAP draft. These are coverage gaps, not evidence of additional production defects.
Remaining scope
- Exercise the real compose Close → Discard path for a pushed IMAP draft: successful admission/dismissal and stale-address refusal that preserves authored content, attachments, and presentation. Reuse existing fixtures; avoid adding production abstractions solely for test access.
- Add a targeted cleanup-removal mutation proving that the existing attachment lifecycle assertions detect missing postcommit cleanup.
Existing evidence to retain
The focused deletion suites already cover native IMAP ownership, pushed-placeholder success/refusal, attachment cleanup and bystander preservation, transaction rollback, and durable retry after provider failure. All five Drafts swipe layouts have passing UI coverage. Earlier mutations detect missing ownership/address guards and the broken swipe action. Do not repeat or broaden that work without a concrete gap.
Acceptance
Add narrowly scoped tests for the remaining cases, with meaningful state assertions and targeted regression detection. Use an isolated simulator only while tests run and shut it down afterward. No draft-delete redesign, legacy stranded-draft cleanup, or unrelated ordinary-mail test expansion.
Why
Track coverage deferred from #143 so that the fix stays focused on Drafts swipe deletion and removing the exactly owned local IMAP draft. These are coverage gaps, not evidence of additional production defects.
Remaining scope
Existing evidence to retain
The focused deletion suites already cover native IMAP ownership, pushed-placeholder success/refusal, attachment cleanup and bystander preservation, transaction rollback, and durable retry after provider failure. All five Drafts swipe layouts have passing UI coverage. Earlier mutations detect missing ownership/address guards and the broken swipe action. Do not repeat or broaden that work without a concrete gap.
Acceptance
Add narrowly scoped tests for the remaining cases, with meaningful state assertions and targeted regression detection. Use an isolated simulator only while tests run and shut it down afterward. No draft-delete redesign, legacy stranded-draft cleanup, or unrelated ordinary-mail test expansion.