If it's not possible to start the main process again (for long enough), there is also the possibility to commit
the container to a new image and run a new container from this image. While this is not the usual best practice workflow, I find it really useful to debug a failing script once in a while.
docker exec -it 6198ef53d943 bash
Error response from daemon: Container 6198ef53d9431a3f38e8b38d7869940f7fb803afac4a2d599812b8e42419c574 is not running
docker commit 6198ef53d943
sha256:ace7ca65e6e3fdb678d9cdfb33a7a165c510e65c3bc28fecb960ac993c37ef33
docker run -it ace7ca65e6e bash
root@72d38a8c787d:/#