Cloud phone farms fail in predictable buckets — not random “bad luck.” This runbook maps GitHub-issue symptoms to fixes before you burn accounts or swap vendors unnecessarily.
Symptom matrix
| Symptom | Likely cause | Action |
|---|---|---|
adb: device offline | TCP session dropped, instance rebooted | adb disconnect → reconnect; see ADB guide |
| Login loop / verify device | IP country jump, new device fingerprint | Sticky mobile IP; do not factory-reset without plan |
| “Emulator detected” | VM-based image, root, bad build props | Upgrade hardware tier; compare vendors |
| 0 views on TikTok posts | Shadow restriction | Pause 48–72h; change content pattern — TikTok guide |
| Shop/seller KYC rejected | Geo/doc mismatch | E-commerce guide |
| App works, web ads fail | Wrong layer — need anti-detect | Meta ads on Multilogin |
Device retirement criteria
- Two consecutive Play Integrity / SafetyNet failures after IP fix
- Permanent action block on primary account after cool-down
- Boot ID or serial churn without vendor explanation
- Subnet-wide ban (>50% accounts on same farm IP range)
Mark device burn in ops log — never reuse for high-value client.
Hybrid stack debug order
- Fix **mobile layer** (cloud phone + mobile IP) for native apps
- Fix **web layer** (Multilogin + sticky resi) for ads/dashboards
- Only then tune automation timing (ADB / Playwright)
Related
Disclosure: MLX-MMO affiliated with Multilogin. Cloud vendor SLAs vary.