I should also add that theoretically folks may still have long-term protection if sufficient memory cells are formed, which cannot be easily be tested with simple serological assays (i.e. the rapid tests). What the study calls into question is the usefulness of serological tests to establish how many folks may have been infected without detection as well as the length of immediate protection.
Now, lack of immunity against RNA viruses in general is often the result of their high mutation rate. OTOH, coronaviruses have a proof-reading enzyme that reduces the mutations rates (but are still high compared to DNA viruses). Also, there a bunch of viruses that can cause cold symptoms so it cannot actually be traced back solely to the major human coronavirus strains.
I have looked a bit into some older pre-SARS papers and found one from 1990 (Callow et al. Eipdemiol. Infect) in which 15 volunteers were infected with coronavirus 229E. Here they showed that some volunteers showed slightly increased antibody titres after one year, though it did not protect from re-infection. However, there was lower shedding, indicating a higher level of neutralization and none developed a cold.
So there is some potential there, especially if vaccines result in a stronger response. At the same time, SARS-CoV-2 (and 1 for that matter) obviously elicit quite different responses, including massive inflammatory responses. So there are still a lot of unknowns at play (plus, we do still do not understand all that goes into long-term immunity and the literature is maddening at best).