I believe this will still fail on $el.replaceWith( function( old ) { return textToBePrefiltered; } ).

@gibson042 could you post an jsfiddle?

Perhaps i didn't get your example, since this -

htmlPrefilter = jQuery.htmlPrefilter

jQuery.htmlPrefilter = function( html ) {
       // Remove <script> and <del> elements
       return htmlPrefilter.apply( this, arguments )
           .replace( /<(script|del)(?=[\s>])[\w\W]*?<\/\1\s*>/ig, "" );
   };

 $("div").replaceWith(function() {
    return "<script>alert(1)<\/script>";
 });

with conjunction with #2203 works for me

Your example works only because the replacer function returns non-empty. I put together a jsFiddle, and actually uncovered (and fixed) a bug in #2206: http://jsfiddle.net/cdhxqwk4/. Observe how this PR erroneously removes all the original elements.

I just want to stop being so damn clever with .replaceWith and get to a point where it does exactly what it says, which I think is achieved by #2206.

@gibson042 this is a lot of code, could you provide simplified example? Just like we ask other people to do?

I just want to stop being so damn clever with .replaceWith and get to a point where it does exactly what it says, which I think is achieved by #2206.

Yeah, your way is more robust and this logic is vague at best, but #2206 complicates a lot in order to fix a very-very edge case.

Just a gist or PR with tests would do

Sorry; I thought the fiddle was a clear comparison between the two PRs that demonstrated errors in this one. Straightforward tests are already at https://github.com/jquery/jquery/pull/2206/files#diff-328a9410239da61336e3662fdbfe9bf6

As for the rest, .replaceWith is full of sharp edges, but I believe that #2206 actually simplifies the logic now, albeit by adding a new parameter to buildFragment and domManip. Trying to guess at possibly-dynamic arguments is just begging for problems, and requiring a this.remove() to clean up after the domManip callback is indicative of bad implementation.

Okay, although those are really weird test-cases, i fully agree with your point.