Hi everyone,

After more testing, this is what I've discovered.

- I can now explicitly reproduce the problem, usually within a minute or two,
  often sooner. [1]
- If I turn off savehist-mode I cannot reproduce the problem.  However, there is
  some other interaction with my configuration that I haven't figured out yet.
  Running emacs -Q and evaluating (setq savehist-mode t), does not trigger the
  problem.
- The problem does exist in 24.5.

I will look into your specific memory-leak debugging tips next, although it may
take me a few days to get back to you.

Thanks for your input,

Joseph

[1]
- open a file
- start a marco
- remove a chunk of XML that just fits in the screen
- add/remove a few blank lines
- yank the chunk back in
- add/remove a few lines blank lines below the chunk
- save the macro
- run the marco repeatedly