Of the four systems I have, 3 of them included manuals - each of those included a printed copy of the MIM (maintenance manual) in addition to the BASIC (mine are BASIC models, so no APL manual). I think that’s evidence that the MIM was fairly accessible. [ edit: that’s for the 5110’s, perhaps might not apply to the 5100 ]
IBM did try to sell “maintenance agreements” or lease options on these systems - but in the 5110 receipt that I have, there is a letter from IBM with a “strong reminder” that basically says “you’ve bought this and are responsible for setting it up.” (edit: exact term was “purchase basis”, basically reminding this isn’t a lease and no formal on-site IBM tech support is included).
And true - in later microcomputers, it was normal to have two categories of books: general BASIC programming books, and then whole separate books about more “system level programming” (re: machine language). So you’re right, Harry maybe just considered it not “need to know” as you say. He was probably near 20 at the time, and now over 80, so hard to say if he has solid recollection about it.
I’m not sure about IBM being “ashamed” of the instruction set. I recall in the S/360 they had built a 1401 emulator also - so they were very keen on the idea of using some new/current instruction set and retaining backwards compatibility with some older instruction set. But there was some internal IBM department struggles about the development - Alvin Ginsburg (one of the SCAMP prototype developers) wrote a letter on 2013 (posted at one of the Wang websites), he described IBM had a thing where one department could write a kind of “stop work order” against another department, claiming some aspect of the product competed with an existing product (in more modern times, Apple and others calls this concept “cannibalizing” existing product lines). My point is, the department involved maybe just wasn’t inclined to really say the specifics of how the system worked - lots of other folks in IBM still wanted to sell the more expensive systems (I’m meaning $100k systems vs $10k systems here). However, it’s not like the 5100 had a full 100% S/360 emulation - from the opcode vector table, we see only about 80% of the instruction set has jump table entries (presumably, they just emulated what they needed to get APL to work).
Dennis did relay to me a story that when the IBM PC came out, IBM did not expect it to be successful at all - and that low expectation was directly related to their experience with the 5100. So IBM had some quite small number of IBM PCs initially in 1981, but quickly had to ramp up production. I wonder if IBM maybe misread the situation – the 5100 may have been more successful than they realized - not in the product itself, but in spreading the idea of how useful such a thing was for even small/medium business, or even just more casual home usage. (I recall reading articles about Altair’s and early Commodore being notorious for somewhat poor quality control - people often had to exchange/return or swap some component {RAM?} about them due to some failure, whereas IBM did maintain a decent quality - and reputation just followed them with the later IBM PC, e.g. they had parity checking even in the keyboard, all the RAM, etc {in the 5100, not so in the IBM PC} – note today that server class IBMs still have ECC memory, or also higher end HP servers too with Xeons).