What has been shown is that if you introduce the SARS-CoV-1 peptide ex vivo into cells drawn from the blood of survivors that there's a IFN-gamma response from T-cells.
That shows that there's clearly some memory in the immune system, but it isn't like they did challenge testing with live SARS-CoV-1 virus to survivors to see if they got sick or not.
And we now know that the presence of a T-cell immune reaction is not sufficient to prevent you from getting sick or hospitalized. The cross-reactive T-cell responses between previously circulating human coronaviruses and SARS-CoV-2 was found to simply not be protective at all against disease:
https://blogs.sciencemag.org/pipeline/archives/2021/02/10/do...
https://www.sciencedirect.com/science/article/pii/S009286742...
So the paper on T-cell responses in SARS-CoV-1 patients 17 years later is not remotely as meaningful as some people suggest. It doesn't prove anything about disease or reinfection.
At the same time the immune response to SARS-CoV-2 actually is likely to be durable for many years as has been seen with HCoV-229E:
https://journals.plos.org/plospathogens/article?id=10.1371/j...
What the researchers there found is solid evidence that our immunity to HCoVs doesn't wane but that the virus mutates in order to reinfect. That contradicts some other studies which seem to show reinfection of ordinary coronaviruses on short timescales, but the methdology of that study seems to be more sound. That's the one that I'd cite as evidence of durable multi-year disease-preventing immunity. Although the catch is that given this viruses ability to mutate in multiple locations around the spike protein while actually increasing affinity it seems very likely that it'll eventually mutate to achieve immune escape. It is looking like it'd be pretty surprising for this to be a one-and-done like measles.