I've confirmed that the problem with dgit vs arm64 infra (large number of distinct autopkgtests combinatorially increasing the chance of failure + looping due to instance allocation failures) is not a regression vs the version of dgit in the release pocket, so accepting this.
I've confirmed that the problem with dgit vs arm64 infra (large number of distinct autopkgtests combinatorially increasing the chance of failure + looping due to instance allocation failures) is not a regression vs the version of dgit in the release pocket, so accepting this.