First off, why it's best to remove it: They are a sensitive device, the temperature is controlled to be in a narrow window for max life span, moisture is the devil for O2 sensors and it's disabled in software and left in the bike moisture can kill it.
Second, as to why it's better turned off: Triumph tunes only use it at low RPM low opening, and they shoot for a 14.3ish AFR, nothing wrong with that to be honest. Read on how narrow and sensors work, they don't measure the afr, they count how many " too rich" or "too lean" incidents it has over a time span and add or remove fuel to compensate. The result is an engine that's never right, instead it swings back and forth across the target lean to rich, but never right.
The range of the swing depends upon how fast the ECU updates the sensor value and does the man. The rocket ECU does both sampling and math pretty slowly, the result is a chunk chunk chunk idle, when you turn it off and set the AFR slightly "too rich" in the mid to high 13:1 range idle will be dead steady and stable.
Additionally if you were to make changes in your F and L tables, and the sensor is turned on but you don't change the AFR target to deactivate closed loop, it'll slowly trim both short and long term trims. The problem with that is that those trims are not limited to the closed loop areas but are instead applied elsewhere in the map too. It effectively undoes any custom ECU work you have.
Add in a PC-V and then there's another layer of computer fighting the ECU which is already fighting the work you've done.
Hope that clarifies your conundrum
PS: The long and short term trim issue is why installing a PC-V using their simulator instead of disabling the sensor in TuneECU is a mistake.