A port can produce, at most 8 bits at a time. That's 2 nybbles. If you have to do something "useful" with those nybbles, yet want to "see" them in 7-segment format, it becomes pretty inconvenient to try to do it with the MCU, particularly in cases where the "visible" form is not a permanent part of the system under study. There's still occasional use for such things.

The industry didn't quit building those BCD-to-7-segment encoders. Few applications demand HEX display, though, and that's probably the reason the HEX-to-5-segment encoders disappeared. Aside from that, there was the SPLD solution, which, while not current-controlled, made it easy to generate a two-digit multiplexed display with that one IC, and many of them were capable of driving both the segments and the common anode or common cathode. Many CMOS SPLD's were rated for 24 mA Iol. With small LED displays,

Nobody would do this if the MCU had otherwise unused resources capable of handling the task, but there's been plenty of discussion indicating that that's not often the case.

Yes, the recollection of the old '9268 encoder was merely a trip down memory lane, but it does offer a solution to a problem for which no comparable solution exists today. The 9368 processed leading-zero conditions and provided a latch for the nybble to be displayed, yet could leave that latch transparent, thereby allowing the device to support multiplexing as well. It had its place.

RE