Another quick bug report: the Bootstrap sampling node introduces ordering by the original RowID after completion, thereby substantially breaking the expected randomness. A workaround is to shuffle afterwards, but having to do so is not optimal. If the implemented sampling strategy introduces an ordering, this order should be broken afterwards (with an opt-out opportunity for those who depend on the not-as-random behaviour).
In case you wonder why this matters: rather than looping 1000 times to create 1000 bootstrap samples for processing, I create bootstrap samples of 100,000 % realtive to the original sample size. I then add a sequential "chunk" id to identify the bootstrap sample in question. This approach is obviously severely distorted by an (unexpected) ordered outcome of "random sampling with replacement". :-)