I stepped out to get a cup of tea and found the IFO unlocked when I returned. The normal lockloss site is hitting an error somewhere early in the analysis, so I ran the analysis on an older version of the code (linked below). There is a glitch seen in ETMX L3 about ten seconds before lockloss, at which point SRC1 and SRC2 start to run away.

https://ldas-jobs.ligo-wa.caltech.edu/~yannick.lecoeuche/index.cgi?event=1244076207