Hi. Currently there is a case when Testcontainers isn't that great. I talk about complicated integration testing environment where there are many containers that are launched during tests. Current architecture of Testcontainers creates needed containers, starts them and destroys them after tests. It's perfectly OK when your tests use one, two, maybe three containers but after that number time to run single test on developer's computer just gets too high.

One of projects I work on has rather complicated integration testing environment: two Cassandra containers, Maria container, Splunk container and several containers with some applications. With Testcontainers creating and destroying all these containers around test execution (one test or suite, doesn't matter) takes ages to execute. Currently the project uses some Groovy scripts trying to find existing containers and use them instead of always creating new ones, but overall experience with this solution isn't good and I'd like to bring Testcontainers into that project. Which has aforementioned time consuming problems with current create-and-destroy approach in complex environments.

So I thought - why don't bring to Testcontainers optional behavior of reusing existing containers (stopped or even running)? In complicated environments it will greatly boost test startup time to seconds - Testcontainers would only need to scan existing containers and if one is found and running - only apply initialScript from PRs https://github.com/testcontainers/testcontainers-java/pull/776/files#diff-5b39417ccdbbe90d4dc26c69fba8fa8eR114 and #575 and container is good to go. If no suitable containers found, then simply create the one - existing Testcontainers behavior.

What do you think about this idea?