Having multiple pipeline instances at same time creates concurrency issues into the database

[lfrias81]

When running multiple pipelines at same time (like in the order of 50-100, each containing 10 jobs), the database encounters concurrency problems and leaves the jobs in running state, though they actually failed because problems with the database were found.

I consider the database should be able to cope with that level of concurrency, but otherwise, the state of the database should at least be keep coherent.