This is a subject I'm very interested in myself, and one that seems not particularly well covered in the various books I've read on cinematography or film style history.
24mm was indeed very wide when Toland shot Citizen Kane, the widest professional cine lens available at the time, as far as I'm aware.
Despite there having been very wide angle lenses since the beginning of the 20th century, including over 90 degree fields of view from lenses designed for large plate architectural photography, and the first fisheye lens being made for cloud studies as far back as 1923, cinematography lenses had particular requirements - more stringent aberration correction and higher resolution for the subsequent big screen enlargement of a smaller negative, tighter tolerances in centration of the smaller glass elements, and faster apertures for the shortened exposure times of motion capture - which in the 1920s limited professional cine lenses to a minimum of 32 or 35mm in focal length, eg Bausch and Lomb had a 32mm Tessar (licensed from Zeiss), Cooke a 35mm Anastigmat.
By the early 30s, Cooke had a 32mm series 0 Opic, forerunner to their Speed Panchros, and Zeiss had a 27mm Tessar. Within a few years Cooke had released their 24mm Speed Pancho, Goerz had a 25mm Kino-Hypar Anastigmat and Bausch and Lomb had their 25mm Baltar. There were no doubt other examples in this range around this time. Certainly by the end of WWII there were several 1" or 25mm lenses available for Eyemos (the most ubiquitous 35mm documentary camera of its age) though this was probably dictated more by the need for wider coverage than the excellence of the image itself. Despite their obvious technological advancement in the field of optics, Germany's Arriflex 35 (the original Arri 2C as used by German WWII combat cameramen) had a lens hood that vignetted on lenses under 28mm. Of course with its newly developed spinning mirror shutter they had a much deeper flange depth to overcome.
After the impact of Toland's deep focus and wide angle style had sunk in, many features were shot almost entirely on 30 to 35mm lenses, including some by Welles himself, but 24mm was still seen by many cinematographers as too distorting, unsuitable for anything but landscapes or wide establishing shots.
Another limiting factor for wide angle lenses for cine cameras (and reflex cameras in general) was the problem of very short back focus distances and the need to clear a reflex mirror or prism. Cooke had already provided a solution back in the 30s when they had modified a 30mm anastigmat to clear the beam splitter prism of Technicolor cameras by utilising a reverse telephoto design that extended the back focus distance much further than the focal length. This opened the way for further short focal length designs for reflex still and cine cameras, something Angenieux began to specialise in by the 1950s with their 'retrofocus' lenses.
According to the Cooke webpage a series II 18mm Speed Panchro was released in 1945, but it was not perhaps fully formed, as cinematographers don't seem to have embraced the Cooke 18mm until the series III version was released in 1954. It became something of a classic, deposing the Angenieux 18.5mm that had briefly reigned after its release in 1951.
Around this time Mike Todd created 70mm Todd-AO as a single lens answer to Cinerama, and utilised a specially designed "bug-eye" lens with a 128 degree field of view to attempt to recreate the 3 camera view of Cinerama. See
In the late 60s the French firm Kinoptik released their 9.8mm Kinoptik Tegea, a fixed focus lens famously used by Kubrick in A Clockwork Orange and The Shining. This was probably the ultra wide angle cine lens of choice for some time, before other manufacturers like Century and Zeiss began perfecting ultra wide lenses, culminating in Zeiss releasing their 8R UltraPrime which blew everything out of the water.
This is something of a potted history of wide angle use in cinema (and also avoided any mention of 16mm or anamorphic or Panavision or standard 65mm), I would heartily welcome any corrections or additions.
Not sure about the first use of an actual fisheye lens rather than just a very distorted wide angle, possibly HAL's viewpoint in 2001?