When working on a piece of code I often get distracted by some badly
named item (function, data type, etc). When I find a better name I
normally use darcs replace to actually do the renaming; or I use replace
--force after doing the change in my editor. In any case, the renaming
is usually something that can and should be recorded separately from
what I wanted to do originally.

Now, if there are many changes in the working tree, it is quite annoying
to have to deselect all the hunk patches (using 'w' key, lest the
replace might be skipped because it depends on one of them) to finally
arrive at the replaces that I want to record.

If replace changes would be offered before hunks, the work-flow would 
become much simpler. After selecting the replace(s), I could say 's' to
skip the rest of the changes. I would expect that in this case (only)
the hunks that depend on the replace are selected automatically and all
others skipped.

The current order in which hunks/replaces are offered might have
advantages I am not aware of, in which case an additional option to swap
the order might be better than to just change the behavior.