Before asking "how best?" to sum analog you should ask yourself if it's actually better? This is of course is subjective and you may only be able to answer it by doing it.
My opinion follows based on personal testing and a lot of conversations...
I live in L.A. and through my work as a mastering engineer and acoustics work I spend a lot of time with high profile mixers. I'm not sure I can name one who is summing analog, either through a console or hardware summing.
Again as part of my work I'll often get the mix from a client in two forms. Just last week I had client send a mix summed in Pro Tools and then the same mix summed through the SSL AWS900, the ITB was better. A few weeks prior it was an API1608 that failed the same test. I literally just came from a friends studio who was telling me how much he's come to hate the sound of his Rupert Neve Designs summing box (fed with Lynx Aurora n).
I haven't mixed a record in maybe 8 years but I did a very deep dive then comparing several summing mixers, ITB, and ITB with 2 bus hardware. I tried every combo and in the end mixed ITB with simple bus analog processing (1 EQ, 1 Comp).
Despite the mythology on Gearslutz (and occasionally here) there is nothing to summing except added distortion. And that distortion can be added very effectively with simple 2 bus gear if it's desired (remembering that it's something of a trade off between the upside of cohesion and the downside of flattening contrast between elements).
Ok, this was a longer answer than I expected! I'm QCing a Bossanova album so had the time I guess