"but there's a spectrum from "the table name is misspelled in the Oracle query" which is a pretty darned safe fix" ...
I wrote my comment above knowing that the fix Benjamin was describing was to simply generate a unique name for a local variable, similar in impact to your table name example.
However ... note my comment about conservatism above. My gut reaction to changes like this is "HEAD, please, not our release branch". On the other hand, this one's extremely safe so I won't object to it going into 5.1 if folks decide that's appropriate.
But in general we shouldn't be putting new features into the release branch once it's feature-frozen, no matter how trivial or minimal they seem to be.
