When they went to Level II basic, they had to fit 3 roms in the space of 2 sockets. No problem, just an add on board. Oops, that board won't fit on the back. We'll tape it to the front and use some long wires. The whole expansion interface part comes across as a total hack.
I get the impression this is what it would have been like if instead of making an Apple II, Jobs and Woz said "well, these keep selling" and just decided to keep hacking away on the Apple I.
That same Apple went on to make the infamously unreliable Apple ///, the Lisa's unreliable "Twiggy" floppy drives, etc.
Steve Jobs himself was responsible for many issues - he insisted upon the Apple ///'s attractive-but-impractical case design (which required splitting the motherboard across two boards connected by a ribbon cable), claimed that any computer with a case fan was junk (see: overheating in the Apple /// and Macintosh), etc.
The earliest home computers (TRS-80 Model I, Apple ][, Commodore PET) used discrete logic for everything other than the microprocessor and RAM. Next generation machines had custom ASICs for the glue logic, display controller and such which brought costs down.
Modern retro-revival computers like the Commander x16 struggle with this because the conventional wisdom is that you need to ship 10,000 units for it to be worth making an ASIC. A display controller, particularly a color display controller, is conceptually simple, a really fun project in discrete logic, but the part count is absurd and since things like the x16 compete with the Raspberry Pi, the cost is a problem. Turns out to be most cost effective to implement a VDP in software with the ESP32