This is an error related to capturing through the org-protocol bookmarklet. Emacs appears to freeze when capturing a huge text selection.
At first I thought it was something related to the error No server buffers remain to edit in the buffer, but I realized that it was still completing captures whose selection length was shorter.
Bookmarklet:
javascript:location.href='org-protocol://capture://w/'+encodeURIComponent(location.href)+'/'+encodeURIComponent(document.title)+'/'+encodeURIComponent(window.getSelection())
Somewhat long website: https://en.wikipedia.org/wiki/Generalized_normal_distribution
Really long website: http://gawker.com/on-smarm-1476594977
To replicate, make sure emacs-server is running, select the entire body text on the page in a natural fashion (mouse etc), then activate the bookmarklet. The Wikipedia page takes a couple of seconds, while the Gawker page takes nearly a minute (and doesn't give any sign that it hasn't locked up).
I do have over 1064 lines in my web capture file, if that fails to replicate it, but I think it's the size of the source page.
How can I speed this up?