It happened sometimes also to me. I suspect that somebody (author? publisher?) could remove the PDF copy of the paper from the repository on the Internet. This is why GS may not see it any more.
If a plagiarism content is detected on the article wherein you were cited, the publishers may remove the paper on the instance of the petitioner without the consent of that author. This is just one major reason.
I experience the same thing but remained unanswered as well. it may be removed by the author for continuation of the project or some privacy issues or consent issue.
I have considered that possibly something caused the Google search to count a paper twice, but then subsequently the error is corrected, and the count goes back down.
When the papers are in "online status" yet, this problems may be emerged. So, you should add your published papers into the researchgate after "open access status".