I often put Terminator nodes on joins that are not expected to fail and TestFilter nodes meant to be exhaustive as kind of assertions to make sure my data makes sense. However, if these assumptions start breaking, they usually break for more than one feature, and if that happens, debugging this once it’s already been pushed to FME Flow becomes a game of whack-a-mole (with the reported error actually hiding a dozen more), unless we modify the flow in order to debug the error.

I do have custom solutions to buffer errors in some cases, but getting this to work is complex and requires a lot of carefully arranged transformers for each potential termnator.

(ProritySwitch is used to ensure no features can come through until all errors have been handled. It’s a custom transformer, more info here)

It would be conveinient if Terminators could be configured to buffer all errors instead of terminating on the first one, to allow outputting and formatting all errors at the same time.