## System details

- **Hardware:** Intel T2 Mac (MacBookPro), Watanare T2 kernel (`linux-t2 6.19.11-arch1-Watanare-T2-2-t2`)
- **Filesystem:** ext4 root on NVMe (`/dev/nvme0n1p3`)
- **Upgrade path:** Omarchy 3.8.5 → 4.0.1 (`omarchy upgrade to quattro`)
- **Boot:** rEFInd/GRUB (stale), transitioned to Limine during upgrade

## What happened

During the quattro upgrade, DNS completely broke mid-pipeline — pacman could not resolve any Omarchy mirrors:

```
:: Checking for remaining system package updates
error: failed retrieving file 'core.db' from stable-mirror.omarchy.org : Could not resolve host: stable-mirror.omarchy.org
error: failed retrieving file 'extra.db' from stable-mirror.omarchy.org : Could not resolve host: stable-mirror.omarchy.org
error: failed to synchronize all databases (failed to retrieve some files)

Upgrade incomplete - do NOT reboot.
```

At the time of failure, `/etc/resolv.conf` was a symlink to `/run/systemd/resolve/stub-resolv.conf` (the standard quattro layout), but `systemd-resolved` was **inactive** — meaning the symlink target did not exist and all DNS resolution was broken.

## Root cause

Migration 1782002156 (`/usr/share/omarchy/migrations/1782002156.sh`) explicitly restarts `systemd-resolved` twice (lines 58 and 88):

```bash
as_root systemctl restart systemd-resolved.service >/dev/null 2>&1 || true
```

The `|| true` means if the restart fails for any reason, the failure is silently swallowed and the pipeline continues. On this system the restart failed silently, leaving resolved inactive for all subsequent steps — including the pacman database sync that follows.

## Fix

```bash
sudo systemctl enable --now systemd-resolved
```

This restored DNS immediately and allowed the remaining upgrade steps to complete.

## Why the restart likely failed

This T2 Mac had several non-standard conditions during the upgrade:
1. The root filesystem is **ext4**, not btrfs — the snapper snapshot steps failed and were manually skipped (related: #6683)
2. The EFI partition was never mounted and had no fstab entry — had to be mounted manually mid-upgrade

Any of these could have caused the `systemctl restart` to fail silently.

## Suggestion

The `|| true` on the resolved restart makes debugging impossible when it fails. Consider either:
- Checking resolved status after the restart and failing loudly if it did not come back up
- Or at minimum logging a warning when the restart returns non-zero

## Recreation

Probably not possible to reliably recreate — it depends on the resolved restart failing silently, which is environment-dependent. A standard install on stable wired/Wi-Fi with btrfs would likely never hit this.