I am battling a similar "self-restarting cursor" . The defect

(a) occurs 100% on one server

(b) occurs 0% on another server

Pseudocode

Declare cursor as..

Open cursor

Fetch next

while @@FetchStatus

do stuff

FetchNext

--** right here, if it hits EndOfData, it just happily fetches First Row of Cursor and rolls on.

end

close cursor

Seems like there is a new bug in SQL Server, and it depends on some unknown aspect of how the server is configured.