TL;DR any URIs encoded via escape() have the potential to throw. escape() encodes them in a format not compatible with the standard decodeURIComponent, and has been deprecated since ~2000.

One potential fix here is to find/replace any escape() encodings that deviate from the standard UTF-8 encodings. From ref-urlencode, it seems to be approximately %80 through %FF.

From the original issue:

Most types of malformed URIs are handled incorrectly, but this one throws an exception:

var uri = new goog.Uri(
    'http://www.cpdee.ufmg.br/~elt/docs/DSP/Microsoft%20PowerPoint' +     
    '%20%20Aula_01_Aritm%E9tica.pdf')

decodeURIComponent blows up trying to decode it.

More concretely decodeURIComponent('%E9') throws.

escape('é') -> "%E9"
encodeURIComponent('é') -> "%C3%A9"