There is a lot of confusion over LEDs and diodes in casual experts between threshold , forward and binned voltage. Some factories test every part and bin them with tolerances for Vf, xy colour coordinates etc and bin every part automatically. Bins are then combined for different customers according to part numbers details. WHite LEDs can have in theory over 100 bins at the factory but reduced to a dozen part numbers by combining bins. ( Some specify there must be 3 adjacent bins for intensity and 3 for wavelength and 2 for voltage. Single wavelength emitters will have fewer variables and thus fewer bins because the White is a balance of the ratio of atomic thin phosphor over a thick blue substrate which affects both colour warmth or XY coordinate, intensity and then there is the ESR variable that affects Vf.
All diodes have a threshold voltage, Vth which is basically the voltage at very dim illumination, like a silicon diode is 0.6V and may be selected at say 1% of the rated current to measure estimate ESR. I usually extrapolate the curve to the axis at rated current to measure ESR.
The Vf is always at rated current (not max) which is exponential but a linear regression to extrapolate Vf is useful. where Vf ~ Vth + I*ESR
The binned voltage is a design and process control by the wafer manufacturer where they buy the wafer and slice and dice into chips, assemble, test and then sort into bins. If they stated 3.2, they obviously didn't state the tolerance as all production has tolerances. but one batch has chips is very controlled can be matched within 1 mV and have a standard deviation, but then gets mixed with other batches that might be different by 10, 50, 100 mV, 200mV or 400mV. The difference is all due to the bulk resistance of the junction or ESR. The threshold is always the same but not useful at low current.
Voltage bins may be sorted into 100mV bins but upon customer agreement, may be 1 bin ,2, 3 or open bins ( more than 6) A manufacturer is free to choose epi-wafers from different sources, stock can get mixed results unless agreement is made with supplier or distributor.
If the packages are not labeled with the min-max voltage, then you have no idea, what you are getting, especially without a spec. (bad form) except based on trust.
The Vth always increases with wavelength with the lowest threshold for IR emitters and highest for UV.
White and Blue have the same because they are the same inside, where some blue energy is converted to a wide spectrum of yellow phosphors to balance the RG reception in RGB with more or less phosphor to make it warm or cool.
Generally the bigger the chip power, the lower the ESR where W*ESR~<1 as a rule of thumb for whites. This will be different for UV and Green etc. and often higher with poor quality parts. A tiny 50mW SMT white chip might be 3.6V with 20 Ohms ESR whereas a 3W white chip could be anywhere from 2.9V ( best) to 3.6V (worst) at 1000mA ( actually 3.6W)
Poor quality parts also have a wide tolerance, which is normal, so they are sold open bin, but binned parts, the factory might scrap or sell the higher ESR or Vf parts to other customers.
I would make sure the Vf tolerance stated on the PO until agreed, if you don't have a spec. Otherwise, for parallel operation in high power, you WILL get thermal runaway with large mismatch.
The other thing in high power diodes is they are always tested with pulses at room temp and will have a negative temperature coefficient (Shockley Effect) which can vary by more than a couple hundred mV.
I have a secret formula for running LEDs with CV sources in parallel operation where matching voltage is critical. The critical point is factored by ESR, and junction ambient thermal resistance and drive current, If a correction is needed, it is often no more than the variation of ESR between devices e.g. a 15 Ohm resistor for 5mm LEDs. which results in very efficient current regulation at the right Vf.