The human aspect

/the scan shows something to make the coroner highly suspicious/

The scan shows a piece of plastic in her stomach. Why would she eat plastic? He retrieves the plastic. It is a note from the victim. Scratched into the plastic is "FENT".

Her body has broken down the drug. Her body does not have access to her scarf covering her nose and mouth. Fentanyl persists in the frozen moisture there.

Just The Scanner:

Given the development of a one micron scanner and nothing else, your drug itself will be undetectable. A proper autopsy will do a lot of chemical analysis to look for drugs. But you have a "perfect" scanner that accurately portrays the physical body down to a bacteria-sized level. It does not, however, tell you anything about chemical composition except by inference.

If we assume advanced computing,

then specific drugs MIGHT leave tell-tale physical changes in the body - or not. We don't know, because we don't have scanners like this, or computers that can look for very fine patterns in those scans. But the scanner itself still can't detect individual chemicals.

If we assume a society where such a scanner exists and fits in,

then it is likely that advanced chemical analysis will be done or at least available well beyond what our current techniques can perform in addition to your micron scanner. Any pre-existing compound is likely to leave specific indicators of what it was and how it killed. But there is always a technological arms race in these things, and it is also likely that compounds exist that people KNOW will mimic others or leave ambiguous traces at best.

SO,

if the murderer is clever, well educated, and deliberate, they could still conceal their crime. If it was a rash act by a desperate criminal, then it is very unlikely they would get away with it unless they are incredibly lucky. In other words, the science can support either conclusion legitimately and have it be plausible. You just need to write the story so it logically concludes the way you want it to.

The scanner is nearly useless.

The poison affected her brain, causing death from hypoxia. But the cells, to a 1-micron resolution, aren't going to look any different from brain cells deprived of oxygen for any reason. The poison might be any inhalant - halothane, nitrous oxide, even pure nitrogen would work - that can interrupt breathing in sufficiently high doses. But I like the nitrogen - I can picture someone climbing down a long narrow chasm and suddenly the villain, in a breathing mask, pours a massive Dewar of liquid nitrogen onto a large heated environmental area at the bottom. The chasm becomes a chimney of death, and she passes out and falls in a way that looks natural on the highest-resolution spy video. The poison leaves behind nothing but air. But there might be some peculiar up and down spikes on temperature sensors, and the villain has a Dewar he writes a month later had leaked due to a faulty valve.

Poisoning

There are substances that metabolize away after death within hours. These can not be detected afterward, not even by their effects. Among them is insulin.

A deadly dose of insulin results in coma and death, but the insulin - a molecule of 5808 Da - itself metabolizes away. Insulin is with that size just in the micrometer diameter size, because it is a complex folded molecule. While a full spanning autopsy can detect it nowadays by detecting the specific metabolization products of the synthetic insulin, they require extensive tests besides the blood test - previous samples and a medical record among others. Further, the metabolization will continue after death, so after about 24 hours even the markers you look for will be gone.

With the right poison, even the initial poison is too small - Hydrogenchloride is 127 pm, which is about 4 magnitudes too small!

Destruction/loss of the flesh

Whatever killed the body, the results of it can't remain for the autopsy.

So if the body decomposes or is dissolved before the autopsy, there is nothing to find. Suppose the body is infested directly during or right after death with a strain of very aggressive maggots or other corpse eaters. These result in pushing the decomposition ahead by a couple of days to weeks from the natural progression from the normal rundown. Initiating infestation by maggots within the first hours and good conditions will pretty much fasten the Fresh Corpse (usually up to 72 horus) and Bloated Corpse (up to 5 days after death) steps into Active Decay within just about one to two days. This usually starts only about 3 to 5 days after death and ends about within a few weeks after the death. If conditions speed it to start within the first 24 hours, quite a lot of evidence will be lost.

In the Active stage and later, all the flesh evidence of a body gets destroyed rather quickly, leading up to skeletonization right after. It is doubtful that any stab wound would survive long into the active decay phase as long as the bones are unaffected and chemical reasons for death (see poisoning) will get masked or even destroyed by the decay quickly. The moment we lose most "soft tissue" evidence we are technically in the Advanced Decay stage, and in this, we rapidly start to lose evidence of the source of death unless the source is preserved in the bones.

So the reason the source of death is unclear is most likely because the corpse was not found quickly and instead spent a minimum of one to two weeks in decay, leaving a partially skeletonized corpse behind.

With tropical conditions and the right set of corpse eaters, even a week would leave a body so skeletonized that the flesh will be of near to no use.

HOWEVER sub-0 °C weather completely halts the breakdown of a corpse and instead slowly converts it into a well preserved ice-mummy over the course of decades. See "Ötzi" or the Incan mummies. On those specimens, a full autopsy could be conducted. Under such arctic environments, a predator eating substantial parts would very much muddy the results: with the cracked open head's contents and the soft tissue from the belly and chest area (heart, liver, kidneys, intestines) eaten by a predator like a polar bear, the actual source of death would become much harder to find, and it even might remove necessary markers like the injection spot.

Can't see the forest for the trees

Just because you have a scanner able to create a perfect replica, and therefore access to all the data that that entails right down to the molecule, doesn't mean that you can make a 100% correct diagnosis of the circumstances of her death.

If you take the equivalent of something that is possible IRL. You have a Windows PC that crashes and you can't figure out why. You also have a device that can copy the PC's hard disk and see the data to a high level of detail, and a debugger which can step through every single instruction the CPU performs and inspect any byte in the computer's memory.

You make your copy and try and boot up and it still crashes. Even though you have access to every single byte and can view every little thing it does, with so much data and instructions running into the millions and billions, it's impossible to tell what's going wrong without an expert to look at a more high level overview, e.g. looking at certain settings, interpreting error messages, looking for missing applications and files, running tests, etc., because there's no individual instruction or byte value that corresponds to "crash".

So your scanner can find which molecules are where, what temperature they are and velocities, etc. and has all the information it needs to make an exact copy. But knowing to diagnose certain symptoms or the presence of certain drugs, etc. is a skill all of its own that needs an expert, who - expert as they are - is still fallible. Because no single molecule, or even a count of those molecules can indicate this drug is present, and while an expert might use the scan as an aid to their diagnosis, they still have to use their own expertise to make it.

The Scanner Was Off.

The scanner was turned off that day. The coroner did not use the molecular scanner. She doesn't know how. It is a specialized piece of equipment.

To do a molecular scan you need to call Scanner Chief Miles O'Brian.

The coroner doesn't get along with O'Brian. He avoids the scanner wherever possible. It is also a hassle to reschedule your work day contingent on the Chief arriving in short order.

This is all above-board since molecular scans are not mandated. The scan is only needed if there is any ambiguity about whether the patient is actually dead.

In this case there was no ambiguity. The patient has been at room temperature for twelve hours.

Little did the coroner know there is a new super drug that gives all the symptoms of death including stalled metabolism and decomposition. The code of best practice will be updated accordingly.

Diagnosis is Heuristic

When you scan for every cell and chemical in the body, you have way to much data to be useful to a human investigator; so, attached to the scanner is software designed to aggregate the data into something diagnostically useful, and AI is notoriously easy to fool once you understand how it works.

When you scan a person there are way too many chemicals present to necessarily be able to identify every possible foreign contaminant; so, for forensic reasons, the AI just IDs the ones that are deemed suspicious according to a database of toxins, and ignores the rest. So, as long as the drug is rare enough, it may get picked up on the scanner, but it will not get flagged as something of note; so, no human operator will ever be informed of its presence.

When the scanner does not identify any known toxins present, all it can do it detect for standard signs of life: Body temperature, heart beat, etc. If a few of the criteria considered necessary to be alive fail, then the software could easily ignore other life signs and declare the person dead.

I think the secret could be that death is really hard to define. If someone's heart is no longer beating, does the machine count that as death? Does it require the electrical charge in the brain to dissipate? If it is the former, it could lead to a situation where someone appears dead to the scanner but is in reality revivable.

With the cold and the cliff, would something like the 'body' freezing near solid but then thawing when the character needs to be revived? The drug could stop the cells from bursting when the person was frozen. I am not sure if it is true or not, but there have apparently been cases of people surviving being frozen more than I would expect. https://www.sciencealert.com/a-woman-famously-survived-being-frozen-solid-40-years-ago-here-s-the-science