It was under way at the turn of the century, so that by the 1840s texts were being written which acknowledge non-rhoticity as normal. I’ve never explored this point in detail, but looking at the books from the period that I have, I see, for example, a clear statement in R G Latham’s The English Langauge (Ch. 1 of Part 3 on pronunciation), who makes it clear that /r/ was on the way out in postvocalic position. After describing initial and medial /r/ as being universally pronounced distinctly, he says ‘At the end… this distinctiness and universality of the sound of r is by no means the case’, and he goes on to say that there is ‘a large percentage of educated speakers’ who make no difference between father and farther , who pronounce cargo without the r , and so on. And he concludes: ‘The rule then stands thus – that when a vowel is followed by r , the r is often dropped altogether, and the vowel made open’. Note the ‘often’. But he later talks of the r being ‘non-existent in the spoken language, being a mere matter of spelling’. I quote from the edition I have (5th, 1862), but the first edition was as early as 1841. I’ve always thought that it took RP a couple of generations to become institutionalized, with other phonemes, such as /h/ and long /a/, attracting more attention as markers of an educated accent (judging by the cartoons in Punch ). Both Ellis and Pitman, born in 1813/14, would have grown up with a great deal of variation around them, so I don’t find it surprising that, as members of the first generation in which RP was being established, and perhaps remembering ‘Pronunciation Walker’ (who transcribes final /r/) they would have kept the /r/ themselves.