great question! I expect quite some people have an opinion about that. Here are my two cents: I’ve worked on some projects replacing Photoshop in applied workflows with open source tools. There are indeed some things which were (or still are?) hard with open-source tools. For example: Slice-by-slice registration of histo-pathological slices can be done manually really well in Photoshop. Also annotating areas showing cancerous tissues with arrows and circles is quite convenient. Similar workflows in Fiji or other open-source tools are sometimes complicated but definitely possible. I also generally warn scientists of Photoshop (usually on slide 2 of our image-analysis courses). Major reasons I see in reproducibility and transparency. To my knowledge, nobody has measured scientifically what Photoshop does to images, e.g. when you drag&drop layers around. Also it might not be known in detail how different versions of Photoshop process images. Last but not least: We cannot have a look in the source code. Thus, we cannot even try to ensure that there are no bugs in core-functionalities.
It all boils down to user training. If they know Fiji&Friends well, they won’t use Photoshop anymore, just because the open-source tools are much more powerful with respect to scientific image analysis than this particular software which clearly has a different purpose.