If I understand the question correctly, it is for cases for long running reports or processes that may take upwards of 5,15,60 etc minutes to run. This program might have some UI in the form of a process indicator e.g. Processing 1 of 1000 records, Processing 2 of 1000 records etc and/or it might have an interrupt button so that the user can stop it. With the typical GDC over SSH setup, you could be assured that as long as the SSH connection stayed up between the GDC on your PC, and the fglrun process on your back up server, then this report/process would continue to be run. The same is true with GBC/GAS/HTTP, that is the report/process would continue to run due to the keep-alives occurring behind the scenes , you are not forced to "implement a kind of job system" by moving from GDC/SSH to GBC/GAS/HTTP.
However you might want to consider the following ...
1. If you decide to enable AUTO_LOGOUT, what will happen to these long running programs with UI that typically take longer than the AUTO_LOGOUT period? ...
2. How difficult is it for a user to close the GDC and what will be their typical response to the dialog that appears, versus the equivalent web behaviour which is closing a browser tab and their typical response to the dialog that appears? I would argue that a typical user is unlikely to attempt to kill gdc.exe, and is likely to respect the dialog that appears if they attempt to do so, whereas in a web environment, closing a browser tab and ignoring the warning dialog, is something they are more likely to do, and it is for that reason you might want to consider to "implement a kind of job system which executes these operations in the background"
Reuben