I'm aware that 40-pins was a high-water mark for dual-inline package chips for a significant time; many CPUs of the early 1980s (8086, Z80, 6800, 6502, etc) used 40-pin packages, but no larger.
Likewise, general purpose I/O chips were frequently in 40-pin packages. Examples I can think of off-hand include Intel's 8255, the MOS 6522, and National Instruments' INS8154. The drive to offer as many pins of I/O as possible on your chip seems perfectly reasonable, given that your competitor would also be doing so.
A number of variants of the 6502 were manufactured in smaller (28-pin) packages with reduced address bus width and/or other changes. This suited the use of 650x chips in embedded devices, where a smaller chip could be more convenient. But I'm not aware of any general-purpose I/O chips that were made available in a reduced package; to my knowledge you'd still have to use a big 40-pin I/O chip alongside your little 28-pin processor.
Were there any general purpose I/O chips manufactured in a smaller DIP package than 40-pin?