I created a minimal example here.
If I click on the input to start typing it selects the node and allows me to type. If I then click on the “images” word under the node the selection goes awry. The bottom copy of the block with an input works differently. If I focus the input and then click on the empty line below it deselects the node and puts the cursor into the correct position.
If you “tab” to focus the button next to the input and do the above the behaviour is the same.