It depends on a lot of factors, but I can speak to the early 20th century - because H. P. Lovecraft was one of the great epistolarians of his age, with a voluminous correspondence. Lovecraft scholar S. T. Joshi estimates he sent 80,000-100,000 pieces of correspondence during his lifetime, from short notes and postcards to letters that could be 50, 70, even 100 pages long in one instance.

For the most part, Lovecraft did not create two copies of his letters; indeed, economy often forced him to use whatever paper was available, including the back pages of stories. This was not an unusual practice; while Lovecraft wrote longhand and Texas pulpster Robert E. Howard typed his letters, they both used this economy measure from time to time.

Also for the most part, we do not have both sides of the conversation. Lovecraft saved several letters - we have an almost-complete run of his correspondence with Robert E. Howard for example - but was not in the habit of saving all of them (not unusual considering both the number of his correspondents). Robert E. Howard likewise preserved Lovecraft's letters carefully, so that the surviving correspondence gives a near-complete record of their relationship-by-mail at the time of their deaths (unfortunately, a number of letters were subsequently accidentally destroyed, leaving only partial transcripts).

One advantage that typing had over longhand was carbon copies; this provided an instant copy of a typescript or letter which could (and did) come in hand for these writers, either to pass around a story without typing it all over, or to re-write a story if the original typescript was lost in the mail. However, there is not much evidence that either Lovecraft or Howard used carbon copies to preserve their correspondence - while they valued the letters from each other, they seemed less concerned about keeping copies for their own records.

Robert E. Howard in particular is also known to have drafted his letters; that is, to write out an initial version of the letter spontaneously and then re-write and type it before sending it on. Four draft letters to H. P. Lovecraft are known to survive in this way (and have been published in The Collected Letters of Robert E. Howard - Index and Addenda), and when compared to the actual letters that have been sent show how Howard changed and developed his language and arguments to better answer Lovecraft. It isn't known how often the Texas pulpster practiced this kind of care with his correspondence, but the initial period of writing letters to Lovecraft was very formal on both their parts, so it's likely Howard was taking extra care and later wrote more extemporaneously.

Even if both sides of the correspondence survive, of course, doesn't mean they are held at the same institution or have been published. A number of Lovecraft's letters, for example, are held in the hands of private collectors and have never been published. August Derleth's letters to Carl Jacobi reside among his papers at Bowling Green's Popular Culture Library, while Jacobi's letters to Derleth reside in the Derleth archive at the Wisconsin Historical Society.

So while we are fortunate to have so many primary resources preserved - and even published; the letters of Lovecraft and Robert E. Howard were published as the two-volume work A Means to Freedom - putting everything together is sometimes still a bit of a scholastic chore, requiring a great deal of effort to hunt down the letters (if they survive) and reconstruct what we can of the correspondence. In some cases where the individual letters don't survive, we can still get an idea of the contents based on other surviving correspondence, filling in the gaps to a degree - but that's more historical detective work.