The legacy restore specs "restore validation and cancellation preserve
live data and allow a valid retry" and "backup and restore recover from
interrupted network requests without losing data" ended with a 5 s
toBeHidden on the restore modal. The modal closes only after the server
answers, so that was a 5 s deadline on a disk-bound restore, which WebKit
CI runners sometimes exceed.
Both specs, and the downloaded-backup spec that already waited on the
response, now share confirmRestoreAndWait: it waits for the
POST /prompt_manager/restore response, asserts a 200 (showing the body on
failure) and the success body, then checks the modal closes.
TESTING.md explains how to reproduce disk-bound timeouts with
--device-write-iops in the pinned image.