This probably applies to early KNIME 5.x versions too, but I’m using KNIME 5.3.3 and whenever I drag the data port out from a node to add a new node, and then type Group By, KNIME seems to think I most likely want Row Aggregator.
As a result, my workflow writing has on a number of occasions gone as follows:
… drag data port… type Group By… press Enter (without looking at dialog)… look in horror at being given the wrong node… select Row Aggregator node… delete… drag port gain, this time concentrate and select GroupBy as the second option on the list
Now I do get that Row Aggregator is the newer, simpler, less “techy/nerdy” version of “group by” but it definitely doesn’t replace it in most of my workflows. I can also understand therefore that if somebody is grouping, then maybe they will want Row Aggregator, BUT…
If I type "group ", “Row Aggregator” doesn’t even get a mention. Yes I should remember to stop there at the word group, or not type the space, but… I don’t. Ever! 
I’m not suggesting that I should be perfect… only that KNIME should be 
On typing "group ", the GroupBy node is presented as the likely thing I want right up until I type the “b” of “By” and then KNIME thinks "aha… he’s probably about to type “Group By” so he clearly no longer means “GroupBy” 
Let’s throw him something newer!
Perhaps KNIME could keep a record of search terms typed, and nodes selected so that over time, it adapts to giving me the node I most likely mean when I type a given sequence of words, even if I just type crazy stuff!