Did this majority-minority distribution originally arise through imperialist conquest, through illegal violence, through government expropriation, through market forces, or in some other way?

Yes, yes, yes, yes, and yes. In fact, one of the great points of obfuscation--exploited to not exactly helpful effect by Louis Changuion's quasi-apologetic Omstrede Land (2011; in English as Disputed Land)--is the actual legality and process of the alienation of the land into the hands of whites or the hands of government, which are similar but not the same thing. This process is something I've written a book and a variety of articles about, so it's in my wheelhouse.

I guess something to start with is how fractured everything is in the era when this process gets its biggest boosts: the British imperial century (so-called) from roughly 1815 to 1914 (well, for us, 1814-1913) as relevant to South Africa. Not only do you have a significant number of different political entities (four major, and varying numbers of minor, entities at given times until 1910 and Union) but you also have a severe disconnect between what the law says and what actually exists.

1814 marks the opening of a new land occupation and ownership system designed to put settlers on the land more densely, with a greater stake in freehold instead of earlier leasehold systems. The idea of this was to put more white settlers on the land in eastern reaches where they could effectively be a standing militia, on land that had been adjudged as Crown property after the elimination or expulsion of Khoe and Xhosa-speaking people following the Fourth Xhosa War (or Frontier War, or War of Dispossession; the older name used the K-word which is an overt slur today). The idea is that most land opened for settlement was gained by right of conquest or because it was 'empty,' a point that plays heavily on myths of Shaka's Zulu state and the so-called 'mfecane' somehow removing people's rights to land. Norman Etherington makes a convincing case in The Great Treks but also in his later article "A Tempest in a Teapot" about land struggles in the Caledon Valley that this mythology served the express purpose of removing prior inhabitants' claims and was used by white, Coloured, and African groups variously to lay claim to territory in the 1830s and 1840s. The Boers, who would pioneer the up-country claim of lands in the 'republics' as 'white,' eventually came out the winners in that particular struggle of competing claims, even though until the 1880s they did not have the power to simply act at will to impose themselves. At the same time, we must remember that this was not exactly some grand conspiracy--those taking up residence on the land did not generally know the conditions underneath the title they bought, or claimed, by right of citizenship or on the market. Often, nobody involved knew the actual circumstances except the people being pushed off that land, but the authorities rarely put much stock in their protests unless they found an advocate to aid them as occasionally did happen.

The British colonies for most of the 19th century (the Cape and Natal) allowed for 'reserve areas' and, for certain Africans of means, the prospect of purchasing quitrent land. The Orange Free State and South African Republic (Transvaal) did not permit this, beyond requiring that land with 'grote stammen' (large 'tribes' meaning ones with significant power) was supposed to be left aside. The inspection and titling of land was, however, deeply corrupt and speculative, meaning that a lot of land was titled away that did not even exist, or that actually lay under large towns and agricultural lands. The Pedi heartland is one of the signal examples of this, as described by Peter Delius in The Land Belongs to Us, but it is only one of many. In effect, the law said that the land was titled to whites and that nobody lived there, but on the ground, people were there. In the areas of the far northern Transvaal where I work, this disconnect existed into the 1940s (and arguably even later). If one gets away from having title to land, and simply who had multigenerational possession of land outside of defined 'native reserves,' the matter gets even murkier--if a white landowner takes over an area and allows a hundred families to remain on a sharecropping basis because he (almost always he) can't possibly farm all that land himself, do they have no right to it? This is why land claims after 1994 were taken from long-term resident farm laborers--many were multigenerational, or at least from the general area.

In the British colonies (Cape and Natal), sometimes chiefs and kings arranged for the sale or trade of land, but of course this does not mean they actually had the power to do it. Law in South Africa, whether British or Boer in political affiliation, recognized certain hierarchies for 'tribal' societies and collective ownership of vague 'tribal' lands vested in (male) authority (see note at bottom), and allowed this--it rarely recognized individuals unless they met (in the Cape Colony) the property and educational tests for voting rights. So the law treated the passage of land as a titled thing in one way, while people actually lived on it in another. The law was built on a fiction that ownership was occupation, so if you got hold of a title and found people on your land, you might have them removed. Ad hoc arrangements were so common, however, that squatters' laws (limiting resident labor to a certain number of people per farm) became a hot-button issue in the 1890s and even after the SA War in 1902. There was a fair bit of fear in rural areas over the 'swart gevaar,' or black peril, manifesting in a spate of black buyers and syndicates snapping up land from impoverished Boers, but this never materialized despite a 1905 court decision (ex parte Tsewu) that allowed black landowners to register in their own names in the Transvaal. Of course, that high court decision would provide part of the impetus to curtail land transfers in the formation of the Union and its early legislation.

(I'm really glossing over a lot here. If things are confusing please ask, and I'll clarify--it's a massive, variable history.)

So the legal body of ownership in the state sits on a series of fictions, and it's this body of deeds that the Union of South Africa as a single entity after 1910 tried to shape into something binding with the Natives Land Bill (later the Land Act, No. 27/1913). Sol Plaatje's Native Life in South Africa (1916) is a classic on the view of an educated African--the first recording secretary of the SANNC / ANC, in fact--on matters in SA during this time. He records that the Land Act was deeply unpopular even with rural Boers and other white landowners across the country, because it upset their individual arrangements with African peasant farmers. What he did not know at that time is that the Land Act would be held to not have force in the Cape Province, because it effectively placed a new barrier upon voting rights by restricting people's power to buy land and thus meet the property test. Until 1936 and the elimination of the last few thousand black voters from the rolls, the Land Act would not have actual force in the Cape. That said, government was fairly generous with exceptions to the Act for people buying land in trust, but mainly because the Land Act itself had restricted people to less than 8% of the land area (increased to 13-14% in 1936, as a concession to the reality on the ground).

So the Land Act locked in a distribution of ownership, roughly, that existed, but the reality was not the same as the legal body of deeds, and that legal body of deeds was often built on fictions as well. Hell, even today people are finding out that they live on land based on a deed that can't actually exist, because some field cornet in the 1870s inspected the same plat twice, and the surveyor had no legal power to correct the deed until a dispute arose--not even under the Survey Act (No. 9) of 1927.

If one had to generalize across the country, it's safest to say that the reasons for de jure claims of land ownership did not match what was actually on the ground, or what was even possible, but it did become aspirational and eventually the South African state did--in most areas--have the power to realize those aspirations. People like Changuion fall back on the 'right of conquest' and the inviolability of the legal deed as a starting point, as though those things are done and can never be redone by anyone else. That is disingenuous at best, because it ignores that law was a technology for colonial (or maybe just 'state') dispossession and expropriation, and without any space for challenging it among most other African societies, it gained much of its authority before anyone tried to reconcile it on the ground. When they did, concessions had to follow in terms of setting up new 'native reserve' areas and the like, but they were always too small to support a growing population. That was usually by design, to force landless people into migrant labor, a fact that comes through clearly in the Cape Colony's own Labour Commission of 1893 and the Glen Grey Papers and reports from 1890 to 1895.

I'd have a hard time suggesting a single work to read, but Van Onselen's The Seed is Mine shows how this vise tightened for one particular long-lived sharecropper. The relevant chapters from the Cambridge History of South Africa, one in volume 1 and one in volume 2, are also really solid overviews with good footnotes. Plaatje's also worth reading, but keep in mind he is trying to write about these matters for a British (and US, surprisingly) public so he turns on his presentation style to reflect his position.

(note: except perhaps for the case of the baLobedu polity under various queens Modjadje, but the form was the same as for kings, so I don't consider it an exception.)