Here are my statistics on my 5 3458a
Production year A3 swap time Calnum Comments
2006 2011 92 Extremely stable, opt 002
1991 1999 1243 Extremely stable, used to serve in a cal lab
1999 Not swapped 19 Extremely stable, used to serve in a cal lab, very good TC
1998 Not swapped 1 Just passes SN18 test, but at 0.4 ppm/day not good enough. Also a bit high TC.
2000 Not swapped 8 Suffers very high TC, around 1 ppm/C, so drift kind of masked by TC, but fair to assume its drifty.
The 1998 and 2000 meters have statistical bias, being drifty at purchase (which I was aware of).
Obviously not very strong statistics with 3 samples, but that means 67% of the stable meters have had the A3 replaced. And the drifty A3 span from 1991 to 2006 here. And seems there are much newer meters that developed (or had from production) drifty A3.
1998 meter looked terrible when i got it. It has serial 2823A20500, so closest brother to one of Tins repair projects
https://xdevs.com/fix/hp3458_u4/. So maybe the "angry hulk" bought two meters to abuse in parallel.... It looks nice now after some refurbishment, but has some A3 drift. Over half a year it has come down from 0.6 ppm/day to just under 0.4 ppm/day. so maybe if it is allowed to run for another year or so it may get acceptable.