cw308,
I have no certain knowledge of what prompted the change. Other cartridges have not. I can speculate that so many cases are manufactured on contract overseas where they may use different numbers, that there may have been cost problems with tooling changes needed to keep the tighter old tolerance. If you look at the CIP drawing for the case, they have tolerances (about -0.008") for the positions of the two ends of the shoulder (junctures with the body and neck¹), but they give no tolerance for case length except in cartridges that headspace on the case mouth (rimless pistol cartridges). So, it may be understood under their standard that the case length given is a maximum, as it is here, and the manufacturers have some leeway on how much shorter it may be. For that reason, some .223 Rem manufactured in CIP countries could be shorter than the old SAAMI standard allowed. If a U.S. maker then goes to one of these manufacturers for a contract purchase, the headstamp die will be changed to the customer's spec, but it is likely cost prohibitive to require changing the cutting and length inspection system for the .223 Rem they already make just to comply with the last 0.010" of length.
RC20 is correct that this is in no way a critical dimension. When the M14/M1A platforms ruled the roost in service rifle matches, it was common knowledge the hard Lake City brass was only good for about 4 reloadings in those weapons before the threat of head separation became serious enough that the brass had to be scrapped. Knowing it grew an average of about 0.010" per loading cycle, many competitors would trim their once-fired brass down 0.040" below maximum trim length so they would only have to fuss with trimming once in the life of each case. Instead of the nominal trim length of 2.005" for that cartridge, they would go for 1.975". Most serious competitors reloading volume back then used the Gracey power trimmer to get the heavy trimming over with. And these were match shooters, so if the extra trimming impacted accuracy very much, they would not have been doing it.
¹ The CIP standard doesn't use a shoulder datum for a basic dimension, as the SAAMI standard does. Rather, they have a basic dimension value, S, from head to shoulder angle origin (apex) that SAAMI does not use.