...i'm sure you realise that high Vf does not mean "suitability" for paralleling
...
No it means "HIGHER suitability for paralleling". If a voltage regulator has Vin regulation stable within 5mV Vout, white LEDs that have a high Vf and soft knee will go zero to full brightness over about 0.4v, so that's 3.0 to 3.4v and paralleling at say 3.3v will be fine.
With an old red LED the range 0-max might be 1.1v to 1.25v, the Vf is much lower and knee is much harder. Running a heap of these in parallel from a Vreg supply with 5mV is not so good.
You need to realise the high Vf and especially the "soft knee" give a very similar result to having a small series resistor inside each LED. manufacturers can take advantage of this, and they do take advantage of it.
You also argue that "nobody knows" if they match LEDs, I'm saying based on common sense they probably do not, because;
1. You can buy LEDs in large batches that have quite closely matching Vf, I test LEDs when I buy large batches. (You can also buy ebay rejects from the manufacturer that have massively varying Vf, sold off cheap to the hobby market).
2. Manufacturers use parallel white LEDs because they CAN because of the high Vf and soft knee, and it saves parts, PCB space, energy etc, all in line wiht making a cheaper product. They don't parallel LEDs to make a product cheaper, but spend unneccessary $$ on labout to match 100 thousand LEDs up!
As far as the soft knee on white LEDs goes, it would not surprise me if this is a deliberate feature built into the silicon, ie a small series resistor. If you test white LEDs some brands have a slightly harder knee and some (usually newer types) have quite a soft knee. A manufacturer that makes LEDs that can be more safely paralleled for the massive market of white lighting would be a preferred supplier to a lighting manufacturer, so common sense again hints there could be something to that idea of deliberately matching Vfs at LED manufature and deliberately designing a particularly soft knee.