I have a process set up with about 30 nodes. I run it with updated data about once a week. I do so typicallly without any problems but sometimes the output has problems. For example for a date it filled in all records as 12/30/13 when in fact the dates were are in 2014. When I delete the node where the error processes are taking place and create again the problem goes away. So it seems like nodes can get corrupted. Any ideas as to why this might be happeneing.
Are you reading the dates from a database? Then it could be a timezone problem. We haven't heard of any problems with corrupted nodes yet.
It does not only happen with dates. For example I run a "group" process and one of the variables used is "ID". The ID are always #'s however after the the group process there are some records that have a dash before the ID (-54357890 when it should be 54357890.
This sounds like a numeric overflow. I suggest you check your incoming data first and also all other external variables your workflow depends on.