I could not find any mentions of these problems online. The article itself has no technical detail.

Looking forward to seeing what the actual problems are. It seems this is the first product to market.

Guesses based off the general subject matter:

• Silica concentrations probably vary depending on the exact position of your head, especially since it’s heavy material. If you mount this sensor even a few meters away from a worker then it’s readings could possibly become invalid, eg because an angle grinder is firing dust a different direction to the sensor.
• Silica is a slang term for a very big category of materials. Some might look completely different to others under certain laser observations, leading to some getting missed (bad) and others materials triggering false positives (leading to the sensor’s screams being ignored by workers).
• Self-cleaning routines might be needed to stop it clogging up, otherwise the sensor starts reporting a higher baseline. They could either choose to report this (“pls clean me” light comes on) or ignore it (bury head in sand mode).
• Alternatively it’s performance might actually be fine, but perhaps it’s still being spruked inappropriately. Government involvement in funding the project might (?) magnify this problem.