Investigate performance differences in environments.

[Macroz]

rems-dev is faster than FD test REMS (about 3x so 100ms vs. 300ms). Could the difference be due to:

• email being sent for real (to all handlers)
• translations are not using a cache
• one handler email is wrong and gives out an error (f1d1f01)
• many more handlers (10-26)
• larger forms (128 fields) (larger test data generated in Optimize form editor #3105)
• event notification integration
• many attachments (esp. answer attachment fields)

Production is maybe even slower (for full cache reload at least, seems not for single events).

[Macroz]

Once #3295 and #3305 are merged and a new release deployed to testing we'll see if this can be closed already.

[Macroz]

If email address is wrong the retrying is stopped. This was in f1d1f01.