When running (very) long time series, we often split up simulations in shorter blocks and link them via final/initial states (mainly to prevent total loss of results in case something happens near the end of the simulation, and to make results more manageable). However, when we schedule such a "chain of simulations" to remote server and also have the results stored remotely, the chain is still interrupted as soon as the local agent from where they were launched is stopped (e.g. because the local pc is shutdown). As the whole chain can take days or weeks to run, this is a major discomfort and takes down the many advantages of remote simulations.
