I really enjoyed the Lava Layer antipattern for incremental refactors that never complete. Having learned to recognize it, I think I'm more aware of the cost/benefit of introducing a new pattern, even if it's better in some way.
That article has changed my behavior in some places as well. Sometimes it's indeed better to sit down and replace the entire old solution, instead of going incrementally. It's a bigger immediate pain, but less following pain.
http://mikehadlow.blogspot.com/2014/12/the-lava-layer-anti-p...