Search for high Early voltage BJTs

What should I look for ?
For a given process, there is an inverse correlation between Hfe and Vaf (e.g. a BC548A will normally have a higher Vaf than BC548C), but other than that, I do not see how you could infer Vaf from modern datasheets; ~50 years ago, manufacturers provided detailed parameters like Hoe, which could give you a lead, but that's no longer the case.

Do not think that opting for a low Hfe type will ipso facto provide a high Vaf: power HV transistors have a low Hfe, AND a ridiculous Vaf.
Also, do not rely on historic datasheets: a modern 2N2222 (or its SMD version) has a much higher Vaf than one made in the sixties
 
For a given process, there is an inverse correlation between Hfe and Vaf (e.g. a BC548A will normally have a higher Vaf than BC548C), but other than that, I do not see how you could infer Vaf from modern datasheets; ~50 years ago, manufacturers provided detailed parameters like Hoe, which could give you a lead, but that's no longer the case.

Do not think that opting for a low Hfe type will ipso facto provide a high Vaf: power HV transistors have a low Hfe, AND a ridiculous Vaf.
Also, do not rely on historic datasheets: a modern 2N2222 (or its SMD version) has a much higher Vaf than one made in the sixties
May be there are some processes giving high VAF ?

VAF is the name for the Early voltage in LTSpice models.
Datasheets do not give VAF, can we trust Spice models ?
I could poke in model libraries to look for high VAF( and BF ) like this library For LTspice users. Libraries of models, examples, etc
 
There are certainly processes yielding a high Vaf, but only the manufacturer has that knowledge.
Vaf has never been explicitly published, even when datasheets comprised 10+ pages.

What historic datasheets included is Hoe (BC107-109 for example), which allows you to compute Vaf, but as I said, this kind of detail is never mentioned presently, and old data is not applicable to modern devices, even if they carry the same type number