Were any of those (few) sold boards ever expanded to the maximum possible RAM? That would be a beast of a machine, for cheap, in 1976.
Also: 65K of memory? Not 64K?
The 1K shift register was used as the 40x24 screen buffer. This is the biggest limitation of that machine; you can't modify the screen directly, you can only push characters on to the end like a serial terminal.
https://www.righto.com/2022/04/inside-apple-1s-shift-registe...
On one hand its annoying having to pump addresses and later rows and columns at memory vs "here's a stream, best of luck using it" but the program complexity and performance gains of random access are very high so RAM wiped out ideas like using CCDs as anything other than video sensors.
Seeing as tech is an endless wheel of everything old is new again (see 70s programing paradigms and stuff like that) "someday" someone will try to disrupt some corner of computing with non random access memory.
The experience would likely be similar to ladder logic programming where all the instructions operate repeatedly and simultaneously, well, at least as a programming model. Another analogy would be spinning rust hard disks, here's a continuous stream of bits pouring off the drive head, now write a program to simulate a block store that uses that bitstream, and that is called hard drive firmware or back in the old days that was a MFM or RLL drive controller.
For cheap? https://jcmit.net/memoryprice.htm has prices in early 1976 of $159 for 4K ($2544 for 64k) and in late 1976 of $250 for 8k ($2000 for 64k)
There also was the extremely cheap $90 for 4k ($1440 for 64k) in between.
I expect all of those are without PCB, and the 4k offerings would require a bigger PCB.
Looking at https://www.in2013dollars.com/us/inflation/1976, you have to multiply those by at least 5 to get today’s prices.