Science operates on a certain level of trust. While principles like transparency about data and detailed, replicable descriptions of methodology can help build reliability into science, you cannot bypass a need for trust because even there you must trust that the data presented are the real, unaltered data, and that the methodology was actually performed as stated. There are rarely enough resources available to check and re-check every claim made.
There are a few ways in which fraud can be revealed, though. Image manipulation is one example. Other examples are statistical (ir)regularities in data that suggest manipulation or fabrication, or p-values or test statistics that are impossible given certain structural aspects of the original data, like points on an integer scale.
Most other types of fabrication are impossible to extract from the published record, though. We all just have to trust. For me, that means that I would definitely extend suspicions of someone known to have a past record far beyond their known problem papers. I cannot comprehend the current hesitancy of journals and institutions when clear fraud is present in published papers, the sort raised on PubPeer and in particular by people like Elisabeth Bik who are dedicated to uncovering discoverable fraud.
The issues these folks raise are not issues of scientific controversy or accidental errors, but in most cases are clear evidence of intentional manipulation. That is, there is no plausible explanation besides fraud. Even if the motivation is claimed to approach innocence like "well, this is representative of what we saw under the microscope, I just did it to make a nicer picture for the journal" or "yeah but the main results of the paper are from a quantitative analysis so these images we photoshopped to show the result don't actually impact the results".
You'll have to decide what level of misconduct has occurred, if any, and act accordingly. If it were me, I'd be gathering up the available evidence and walking directly to my advisor's office and confronting them. At that point, I wouldn't be concerned with our relationship: if there are credible accusations that they cannot defend in a meeting, I would not feel comfortable having a professional relationship with them.
That said, I mentioned that journals and institutions have been hesitant, especially when involving powerful or notable people (e.g., Nobel prize winners and directors of large research programs). Punishments, when present, have often been silly, like brief (1-2 year) bans on research funding, or a stern warning along with some retractions. These punishments do not fit the offenses committed. I hope this will change in the future, but in the meantime, it sadly appears that a profitable academic strategy is to have one of these fraudsters promote your work within the community of people that for some reason still trust them. Maybe you'll get hired to work with their previous advisor who trained them to cheat, or with their former student who also benefited from some deceptively beautiful papers in fancy journals. I wouldn't feel comfortable with that, but I also recognize that blowing up your current PhD training to go elsewhere is going to cause you harm in the short term (if nothing else, by some lost time) and that your actions alone will not be sufficient to change the system overnight.