I've set up a situation where:
* changing a field triggers a time-based WF rule, which after an hour
* changes another field on the same object, which
* triggers an apex trigger, which
* changes yet another field

This works fine in general, for small record sets.

However, I just tested it in bulk (by making that first field update to a bunch of records at once), and it failed governor limits (too many DML rows).  I didn't expect this, because as far as we the developers can tell, each record is put on the time-workflow queue separately.  However, it appears that since I triggered the time WF at the same time for all the records, they all came off the queue at the same time and were batched for trigger processing.

Specifcally, I tried this operation w/ 380 records.  It looks like it did one batch of 200 and one batch of 180.  So exactly what you'd expect if there was no time WF involved.

But the weird part is: it seems to have batched the records, but not scaled the governor limits to account for it.  So in my batch of 200, I got an error message saying: too many DML rows 201.  But for a trigger operating on 200 records, the limit for DML rows should be 20k.  And I got 200 separate email error messages delivered. So it seems to be treating it as partly a batch, but partly 200 separate operations.

Can someone help with this?  Is this how it's meant to work?

Thanks much!