claims 25% variability to be acceptable
That seems to be quoted fairly often as an industry standard. I suspect it is an artifact from the days of 10 and 20% tolerance resistors and -20+50% tolerance caps. With that kind of tolerance in the rest of the circuit it didn't matter as much if the tubes were a perfect match or not, since you would still probably have to make adjustments elsewhere in the circuit to match tubes up. Not that tubes are as easy to make as resistors these days, I'm sure it is much more difficult to get a consistent tolerance in some aspects of the construction. In the days when tubes ruled the high volume production meant that the really out of spec stuff could just go to the crusher. I can't imagine the demand for 2A3s being high enough to allow a manufacturer to replicate the legend of WE smashing all but 25% of their 300B production to maintain a quality standard*
There are tube vendors who charge a little extra for tubes that are matched to something like 5%. The 2A3s and 300Bs we get from New Sensor are usually very nicely matched. The 2A3 tubes we used to get from Valve Art we had to match ourselves, and every once in a while a couple out of a shipment would simply end up in the dumpster as unusable. Of course we payed a heckuva lot less than $239 for them.
* and perhaps it is an urban legend in the case of WE. I was told that when the 90s version of the WE300B was first prototyped one of the people who worked on the original 300B production said that the new one wouldn't sound the same because the equipment used pulled
too hard a vacuum in the envelope!
