"Simplest possible" is fine until you find that the stand out nicest system was designed exactly for your simplest case but completely fails to allow you to do some of the non-trivial stuff required by your real application. Sure, you can get a feel for how the tools work by writing a Hello World application using each tool, but that is no where near enough to actually evaluate their performance over the long haul for real tasks.

To avoid writing a major application using each of the candidate tools a good technique is to ask other people for their opinion then filter the candidate tools down to a small number for serious assessment.

Graciously acknowledging what you say, I submit that both of our only-slightly contrasting points of view are valid.   Use other people’s opinions perhaps to rank the packages in descending order of likelihood, then get your feet wet with those packages which look-n-feel the best.   Then, as we both seem to say, you’ve got to get into the “non-trivial stuff required by your, real application.”   Whether or not you start with a friendly Hello, you do need, I think, to make a serious attempt to “evaluate their performance over the long haul for real tasks” before committing to a final decision.   Because, of course, that’s going to be a marriage:   you’re going to wake up in the morning looking at your choice for a long time to come.