I had heard that the 8-bit versions were just a way for Intel to sell some of the 16-bit versions with partial errors. If only half the output path was working, they could sell it as the 8088/80188 chip, and still recover some of the cost of making it. And get some business that needed a cheaper chip, and might have gone to another manufacturer. Chip-making was rather iffy in those days, and there were often chips that failed the testing as they came out of the foundry.
In the newest chips still have such problems, because of the tiny size of traces nowadays – the tiniest flay in the substrate will make parts of the chip non-working. So now the chips are often designed with ‘extra’ parts built-in, that can be brought on line as substitutes, to make sure that chip still functions up to specs.