What historical reason is there for these instructions being called that?
"Historical Reason" is the right key word here, as ...
TL;DR: It's Piled Up Heritage
The 8080 inherited the mnemonic names from the 8008, which in turn is based on the Datapoint 2200 designed in 1969, way before terminology settled. From there on, it was a combination of keeping known names and adding new ones without violating design restrictions. It was up to Mr. Stephen Morse who did the instruction level design of the 8086 to straighten naming and adjust it to by-1976/77-settled terminology.
Terminology is a constant problem when looking into the past (*1). It needs time to settle into (mostly) canonical form. Words/phrases from before that consensus is reached may not always fit today's way of thinking. Likewise what seems settled today may sound wrong a few years from now.
The Long Read:
Intel neither invented nor named these instructions for the 8080, but carried them over from the 8008, which already featured all 4 of them. The 8008 in turn is the single chip implementation of Datapoint's 2200 terminal CPU; thus Intel inherited most names and mnemonics.
The Datapoint 2200 featured two of the four, RRC and RLC, or as they called them SRC and SLC:
(Taken from the August 1971 "Datapoint 2020 PROGRAMMERS' MANUAL")
Being designed in 1969/70, the 2200's CPU predates any modern microprocessor as well as today's canon of assumptions when trying to make sense of mnemonics. (*1)
They called 'Shift Circular' what today might commonly be called 'Rotate'. 'Shift Right (or left) Circular' follows more natural English grammar than 'Shift Circular Right', doesn't it? These instructions were mostly intended to allow serialization of bytes and bit tests, due to being fast single byte instructions with a direct testable result in carry.
When Intel created the 8008 in 1970/71, they added two more instructions, now rotating through carry, allowing code to 'carry' shifting/rotating operations over multi byte values. At the same time they changed language from 'Shift' to 'Rotate'.
(Taken from the "8008 8-BIT PARALLEL (*2) CENTRAL PROCESSING UNIT USERS MANUAL" Rev.2 of November 1972)
SRC/SLC became RRC/RLC. The new one became RAR/RAL. Unfortunately they no longer gave a plain text translation for the mnemonic. The descriptive text is at best hinting. What seems clear is that RRC/RLC are continuations of Datapoint's terminology, easing the switch from that. 'C' was therefore still assigned as 'Cyclic', so it could not mean Carry. Intel adhered to the carried-over meaning where other manufacturers had the freedom of clean sheet designs.
For the new ones, RAR/RAL, they had more freedom. Except, it was bound by
- Using 3 and only 3 letters
- Start with 'R' as they are now 'Rotates'
- Include the direction
- Not using a single 'C' as that was reserved for 'Cyclic'
Solutions that would add another 'C' for carry would invalidate #1, as it would need a 4th character for the direction (RRCC/RLCC) (*3). In the end, only one letter was replaceable, which was the C. They did go with an A, which may mean Accumulator, but as seen, it's not spelled out.
Another reason may have been orientation on existing/known machines. At that time (~1970) the DEC's PDP-8 line (and its offspring) was rather well known, which used the RAL/RAR mnemonics for Rotating Accumulator Left/Right. It was explicitly named that way. Except for the PDP-8 handling a 12 bit accumulator, that is.
Our Hero 8080
With the 8080, the instructions became renumbered, but mnemonics were kept. A dedicated cleartext meaning was added in the Programmer's Manual, one not really resembling the mnemonic:
(Taken from the "Intel 8080 Assembly Language Programming Manual" Rev.B of 1975)
The Great Reset 8086
Last, when the 8086 mnemonics were done by Mr. Stephen Morse, the whole set was overhauled, dropping most of the accumulated ballast and renaming them in a more consistent manner and rather close to what was canonical terminology by 1976/77.
Ignorance is Bliss
In the end it doesn't matter if a mnemonic makes sense to everyone or anyone at all. Its sole purpose is to uniquely identify a certain instruction/working, without writing a novel. It may be nice to 'read' them meaningfully, maybe even helpful, but that's lower priority. Thinking too much about that just ties up even more brain cells without anything to gain ... a straight way to depression :)
*1 - An issue that's quite obvious here on RC:SE, where many questions as well as answers run into the trap of assuming that terminology is consistent over time - or even fit to describe items before it has been coined.
*2 - Interesting side-note: Intel explicitly advertised the 8008 as a parallel CPU, which unlike today isn't a hint about parallel processing, but the fact that it's not a serial implementation.
Another example of a shift in common terminology, isn't it?
*3 - Zilog did go that way, albeit by completely breaking the 8008/Datapoint heritage, naming them RLA/RRA and RLCA/RRCA, creating a quite logical system:
- Rx -> Rotate Right/Left register or memory (*4)
- RxA -> Rotate L/R Accumulator
- RxC -> Rotate L/R Carry and register or memory (*4)
- RxCA -> Rotating L/R Carry and Accumulator.
Zilog also added shifts, named as SxA, where A finally stands for Arithmetic.
Redesigning the mnemonics was eventually one of the most beneficial additions Zilog did - so good that even direct 8080 descendants, like the Gameboy's LR35902, can be rightfully described as a Z80 "inspired" CPU.
*4 - Since Rx and RxC also work with A, this may seem like a renaming of RxA/RxAC, but both are really new, slower opcodes in addition to the ones inherited from the 8080.