We have recently identified a bug/issue on the OAI-PMH interface of OJS 2.4.x (tested in 2.4.6 and 2.4.8), related to the duplication of OAI items in different resumptionTokens.
This issue has been identified because we harvest on a search portal many OJS journals, and after some accurate analysis of the items of the journal, the active items on the OAI and the harvested items, we found some differences.
Also, the error is not constant, but random. The duplicate items are sometimes different and appears on different resumptions tokens. In one test, the duplicate items appears on the first and last resumptionTokens. Also, this didn’t affect all journals of the same OJS installation.
We didn’t find any errors on the log files.
To exemplify, we have listed the first 400 items of the OAI-PMH interface and put everything on a spreadsheet and identified the duplicates on these two tests.
On the first one, we identify 2 duplicate items (oai:ojs.revistas.rcaap.pt:article/1269 and oai:ojs.revistas.rcaap.pt:article/2644), on the second test, wie found these 2 but with many more duplicates. Find here the 2 files for the 2 tests and the duplicates in red on the last sheet: https://drive.google.com/file/d/19Vlnxd6429DXNZ7Ellb525p6qQQSCqkV/view?usp=sharing
So it seems that the creation of the OAI is based on some kind of random list or cache. Do you have any idea of what can be causing this problem?