Since when and why do some academic publishers use, prefer or enforce endnotes instead of footnotes

The main reason for publishers preferring endnotes over footnotes is financial:

Many university presses now more or less require endnotes, since typesetting notes at the bottom of the page requires more fiddling by technicians and is therefore more expensive. Footnotes also carry the potential for added expense when corrections are made to page proofs, since even minor changes can launch a cascading mess, bumping note callouts to different pages and dragging their linked notes with them.

The 'cascading mess' referred is very real as I remember this problem when writing my PhD in the early 1990s. We were required to use footnotes and it was sometimes a nightmare when editing drafts. Texts go through several edits, sometimes adding, sometimes removing, sometimes rearranging - this last one was particularly problematic for footnotes. Also, you'd end up with tables which previously fit on a page being split between two pages or the titles of graphs appearing on one page and the graph itself on the next page.

Academic publishers

have been using footnotes regularly for about a century. Footnotes have existed for probably a couple of centuries or more but they became more regularly and widely used through the middle of the 20th century.

The use of footnotes in academic literature was still widespread in the late 1980s, but the endnote was gaining ground. The site Historiann says

...footnote-killing is a longstanding trend among non-virtual academic book publishers for at least twenty years. Most university presses and tradey U-press lines use endnotes, period.

The author of this source mentions being told that the increased cost of paper was a factor but he/she seems unclear how endnotes save paper. This could be true, though, due to the margins required on each page between the main text and the footnotes. Also, long footnotes can be difficult to deal with and make economical formatting (i.e. using the whole page) difficult, especially if the text contains images, tables, graphs etc.

Endnotes are indeed a pain for the reader, especially when using pdf files, but some Kindle books now have pop-up footnotes (just google kindle pop-up footnotes for more details).

I fairly often have to produce decently formatted printed documents. Endnotes, which I too dislike, are much easier to manage for the publisher. Footnotes on the other hand mess up your page layout big time.

If they designer of the layout complains loud enough - and those prima donnas do that very well - the publisher happily gives in.

To give you an practical example: National Geographic often has beautiful but almost useless colored graphs in articles. For example the population of an area in shades of the same color.

The artists who produce them are invariably young people with 20/20 eyesight. The editors who approve them are slightly older people with almost the same eyesight. But ... most of their readers are old fogies like me over 50 years. They have lousy eyesight and get slightly colorblind. We can't see all those darned shades.

However, as long as the readers don't complain, and the editors don't care, NG will keep those useless but colorful graphs. It's known fact men >50 can see less color variations. Problem is that nobody does anything with this known fact.

The same applies to footnote/endnotes. I prefer footnotes, even though it can screw up your graphic design. Editors look at the cost, and stick to endnotes. The editor controls the purse, so what the editor says, goes!

Short story: endnotes are cheaper and easier to manage.

It is a bit complicated...

Since printing exists, its applications created professionals who were creating documents by hand. Books, and in later times magazines, were expensive because of the necessary amount of manual work involved. On the other hand people used handwriting for their own works; even dissertations were written by hand until the 1970s, especially in mathematics.

So you have a division: The professional setting with expensive books and all other works with handwriting.

The professional work indeed eschewed endnotes for longer works, instead they used marginalia (which seem now extinct) and footnotes. Only articles in scientific journals used primarily endnotes because it was only a few pages and therefore easily accessible. The professionals also used every typography feature available: Kerning, ligatures, removing widows and orphans. The mathematical journals also developed their own tools to layout mathematical formulas. All other works were written by hand.

Then came the typewriter.
It enabled an increase in speed, but it did not support any typographical features. People used it, but began to use endnotes because footnotes were difficult to fix and looked like crap. If you look for mathematical literature from the end of the 19th century on, you will find copied typewriter books with manually inserted formulas.
It looks extremely awful. So awful that mathematicians even then preferred to write their dissertations by hand.

Fast forward, 1970s. Donald Knuth developed TeX which allowed users to develop professional looking documents with full footnote and mathematical symbol support. It is still the standard in the mathematical and technical community. On the other hand, there are those who were using Wordstar (old forgotten standard), Word and OpenOffice, who are still unable to correctly layout footnotes.

While the TeX users never had a problem with footnotes, the latter users preferred endnotes (because their programs are broken). Now the availability of powerful hardware, excellent printing capabilities and more and more content available on the Web and in digital formats led to a decline in the printing industry. Unable to be paid for high quality, printing is now more and more done by unprofessionals.

The claim that footnotes are expensive and error-prone is ridiculous; since TeX it is a solved problem. The problem is that the printing houses are not able to maintain quality, so they are now paying people who only know the Word-style endnote solution.