I was taking a look into this too. Since there was a 1 second gap in the @OgNasty -> @seoincorporation example in the OP, grouping by exact time would not catch these cases, so I went on to giving a 2 second margin (searching for all TXs within a 2 second gap with the same implied users and message and date).

I make it 317 TXs historically by the above criteria (see https://docs.google.com/spreadsheets/d/1JmMF_SZX0oWgK8L2BL-lkB0aDMAIW7HB2OWJKXxtrag/edit?usp=sharing), with 155 different real intended TX sends (the others are side effects of double or triple clicking the send button).

No too many cases, and roughly let’s say half of the implied sMerits being non-intentional (602 sMerits -> would mean 301 non-intentional sent sMerits historically).