Let's consider this from the perspective of the question How can I determine what records are available in a particular locale? Only a small subset of historical records are online -- before that can happen, a lot of other things have to be done first.
How are online records created?
- Records are created, most often because of some law requiring them or some need (e.g. school admissions registers)
- The agency keeps them for some period of time, then discards them or sends them to an archive
- The archive accessions the records and prepares them so that researchers can access them
- The archive microfilms the paper records or makes digital images of them
- The records are leased to an online provider
- The archive or data provider prepares born-electronic records for publication online and / or (optional) indexes paper records so they can be searched
- The data provider publishes the collection so their users / subscribers can view it
What happens next?
Case 1: big archives
Let's assume that a collection is fragile and the records are filmed or digital images are made so they don't have to be handled again. My understanding is that the archive in question will make the images available to users who visit. Most of the agreements I've seen that involve collections being published on Ancestry have some provision that the agency providing the records will also have a copy of those records for on-site use.
Users can check the availability of records offline by looking at the original source of an online collection and doing a search for the microfilm roll or book or whatever the data was.
The original agency may refer you to the online source, but as far as I know, archives and other agencies (for example, the US National Archives, or the county record offices in England and Wales) will provide access to the online service at their offices.
Case 2: local societies
A local society has produced a set of local records (e.g. a cemetery transcription) and has gotten tired of selling booklets or the booklets have gone out of print. Let's say they decide to turn the data over to Ancestry and it gets published there. I would guess that most publications of this kind would still be available in local public libraries, and repositories like the Allen County Public Library, which has one of the biggest genealogical collections in the world. The licensing to put the collection online does not usually require the society to round up all the previously-sold printed material and dispose of the printed copies -- that would be impossible.
Case 3: But what about Germany?
There are some projects which are being undertaken to put church record books online in Germany. Naturally it will take a lot of resources to make this happen, and it's not uncommon for an agency to make records available as pay-per-view. Naturally they want to get back the costs of putting the site up. So the question is, is there a tendency to restrict access to these records so that only people who pay can see them?
It may seem in a case like this that records have disappeared behind a paywall. However, I think it's important to consider what would happen if I went to any archive and asked to take away a copy of a record with me. Not all archives allow patrons make their own copy of a record -- every archive has different rules. Because of the condition of the records, it may only be possible for the staff to copy the records for you. Usually there is a copying cost involved when you ask for such a file, and the fees aren't cheap. But I don't consider that to be the same case as having records behind a paywall, in the sense that many people might (i.e. "You have to pay Ancestry or you can't see these!".)