One may want to run a job, connect to it with e.g. telnet console to examine engine state etc., and be able to do it even after the actual crawl has finished, so that the process didn't exit and the engine etc. didn't finalize and close everything.
It may be weird or not possible to implement cleanly, as important code runs on finalization (e.g. flushing feed exporters) and in this mode it's not clear who and when should initialize this, but maybe the non-ideal solution will be useful anyway.
One may want to run a job, connect to it with e.g. telnet console to examine engine state etc., and be able to do it even after the actual crawl has finished, so that the process didn't exit and the engine etc. didn't finalize and close everything.
It may be weird or not possible to implement cleanly, as important code runs on finalization (e.g. flushing feed exporters) and in this mode it's not clear who and when should initialize this, but maybe the non-ideal solution will be useful anyway.