I would say that admission to Oxbridge has always been ‘meritocratic’ - it is the (often unconscious) factors taken into account in the measure of ‘merit’ that has varied over time.
Long ago, merit was measured through previous schooling, personal recommendation and some element of exam / interview.
Then it was measured through exam (taken often in an additional school term, not offered by many state schools) and interview.
Then through an evolving balance of prior attainment, grade predictions, interviews and subject-specific tests.
None are wholly bias-free measures of merit. What each does is select a slightly different cohort from the pool of ‘admittable candidates, which is much larger than the number of places.
It’s interesting that when auditions for certain instrumental music posts became genuinely ‘blind’ (numbered candidates played behind a curtain) the results were different in terms of gender (and iirc race) from when the auditioners could see the candidate, despite the fact that auditions were always ‘on merit’.
So I genuinely don’t think there is a shift from ‘not on merit’ to ‘on merit’ - in the sense that ‘merit’ has always included subjective judgement of eg accent, confidence, articulacy and unconscious bias in terms of ‘like / unlike current successful students’ - just a shift in how the different elements of ‘merit’ are scored and balanced. If in 2015/16 the shift was slightly too far in terms of allowance for prior education, the 2019 analysis can inform the 2019-2021 admissions, and we may see a change in the coming years of results (or not, as Covid may have introduced a further variable).