Recommended approach for AR tracking of split AC indoor units — low-texture, near-symmetric, near-identical SKU variants

What we are building
--------------------
AR service and self-help features inside an existing consumer mobile app for a
major air conditioner manufacturer. We are evaluating Vuforia Engine and would
like your recommendation on the right approach before we commit to a design.

The objects
-----------
Wall-mounted split AC indoor units.

- Elongated rectangular body, approx. 800-1100 x 300 x 200 mm
- Glossy white plastic fascia, very low surface texture
- Low geometric complexity, few sharp edges or distinguishing features
- Near-symmetric about the vertical axis
- Multiple SKUs across the range are near-identical in external appearance
- Hinged front intake flap (moving part)
- A rating plate with model number and barcode is present on every unit

Usage conditions
----------------
- Mounted 2.5-3 m above floor; users view upward from 1.5-3 m away
- Typical indoor residential lighting, often uneven or backlit
- Two user types: consumers (standing in the room) and service technicians
  (close range, on a ladder, unit panel opened)

Constraints
-----------
- Native SDK integration into existing Android (Kotlin) and iOS (Swift) apps,
  not Unity
- Minimum Android 10 / iOS 14, mid-range consumer devices
- Must work on units already installed in customers' homes, so we cannot rely
  on markers being applied at manufacture
- CAD models are available from the manufacturer
- Deployment would be on a Premium plan

What we would like from you
---------------------------
1. Which Vuforia features would you recommend for this object class, and would
   you consider these units suitable targets at all?
2. How would you approach distinguishing SKUs that look near-identical?
3. Where would you expect this approach to fail, and what would you recommend
   as a fallback?

Thanks