The IceCube Neutrino Observatory, located in the deep ice at the
South Pole, is designed to observe neutrinos above 1 TeV; however, it is also
highly sensitive to low-energy neutrinos from a Galactic Core Collapse
Supernova (CCSN). SNDAQ, an online data acquisition and trigger system designed
to observe CCSNe neutrino bursts in real time, is running with 99% uptime. In
its current implementation, the time windows used by the SNDAQ trigger are
tuned to typical supernova simulations and the observed signal from SN1987A.
However, no galactic supernovae have been observed with high neutrino
statistics so far, and many simulations do not produce an actual explosion.
Therefore, it is wise to define a trigger that is not biased by simulations. To
improve the sensitivity of the trigger to a much wider range of models, as well
as unusual hadronic physics or physics beyond the Standard Model, we have
implemented a time-domain search using the Bayesian Blocks algorithm. This
technique allows the data themselves to determine the natural timescale of
excess counts above background. The Bayesian Blocks window makes the SNDAQ
trigger more robust to uncertainties in CCSN neutrino emission models. In
addition, it also allows for general sub-threshold transient searches. We
describe the implementation and performance of the Bayesian Blocks trigger and
discuss improvements in the sensitivity of IceCube to supernovae in the Galaxy
and its nearest satellites.