Decentralized, cluster-managed deployment of software updates in a multi-cluster environment

Generally described, one or more aspects of the present application relate to deploying a software update to multiple clusters in a decentralized manner, where the individual clusters manage their own deployments based on successes and/or failures experienced by other clusters that have installed the software update. For example, a set of least conservative clusters may install the software update right away, whereas a set of more conservative clusters may wait and see how many of the clusters have installed the update so far and/or how the update is performing on those clusters, before installing the update themselves.