Github user mosermw commented on the issue:
https://github.com/apache/nifi/pull/2767
@mattyb149 I think the issue is whether there is a reasonable expectation
that a user would loop back a 'failure' relationship, and precedent set in
other similar processors. For example, I consider ReplaceText in the same
category as processors that modify content such as CompressContent,
UnpackContent, and EncryptContent. None of those processors penalize flowfiles
sent to failure. In those processors it's not reasonable to expect a failure
to correct itself, so it's not reasonable to loop back the failure
relationship. I just followed that precedent when modifying ReplaceText for
this PR.
---
