Function migration_failed_dialog_body 

pub fn migration_failed_dialog_body(reason: &str) -> String

Body for that dialog. Kept here (not in each supervisor) so Windows and Unix show the same words.

Every sentence is something the launcher actually knows:

• Restart-to-retry: srv re-runs every unapplied migration on the next launch (run_pending_migrations filters on migration_is_applied).
• The backup ordering: runner.rs::run_pending_migrations calls backup_stores (→ ~/.agentmux/shared/backups/pre-migration-<ver>-<ts>/) BEFORE applying anything, and returns Err("... backup failed: ...") without running a migration if that copy fails. So the reason text itself tells the user which case they are in; we state the ordering and let it do so, rather than asserting a backup exists.

We deliberately do NOT claim “your data has not been changed” or “a snapshot was taken”: the backup step is one of the things that can be the failure, and bootstrap/stores.rs’s separate maybe_snapshot_pre_migration is best-effort (non-fatal on error, Ok(None) when it decides none is needed). An earlier revision of this text made that claim unconditionally (#3043).