Also in terms of mistakes codereviewdoctor twice linked to the same issue in their blog https://github.com/tensorflow/tensorflow/issues/53636 and raised the PR to the wrong project https://github.com/tensorflow/tensorflow/pull/53637 (I guess Tensorflow vendors Keras, easy mistake)
STOP!
This folder contains the legacy Keras code which is stale and about to be deleted. The current Keras code lives in github/keras-team/keras.
Please do not use the code from this folder.
The fact they didn't find the same mistake(s) in keras-team/keras (I assume they scanned, it's one of the most popular Python repo) makes me believe these issues have been fixed/removed in up-to-date karas repo.
https://github.com/keras-team/keras/issues/15854
resulting in
"Absent comma results in unwatned string concatenation on line 330"
Bug-ception!
Also a factor that bugs in functional code are more visible, both during development and to users once shipped. So there may have been an equal number or more such bugs in the non-test code, that just didn't remain in the code base for this long.