What concrete steps can be taken / avoided to convey that the invention is not obvious?
I would avoid writing that you used "existing technology". Most inventions are improvements on something that previously existed. As you mention "unexpected result" is one key to not being found obvious. There is also "obvious to try". If there are manageable hand full of different ways to put two things together then you will see that argument from the examiner. One way for you to approach this is to assume you will get an obviousness rejection and understand the ways one fights an obvious rejection.
One way, for example, is to argue that the base reference, when combined with a second reference, is no longer useful for its original purpose and that it's operating mechanism is defeated. A good argument is that there is an element of your claim that is not found in any of the individual references. This is called the "all elements rule". If you claim requires an A, a B a C and a D, the there needs to be an A, a B a C and a D within the multiple references cited in to 103 rejection. And also, "inventions" do not get rejected, specific claim wording is what gets examined and objected to/rejected/allowed.
Another approach to a 103 rejection is to argue that one or more of the references are not from an analogous field. That is a hard argument to win. More practically, you often need to amend a claim to get out from under an obviousness rejection, so you want enough support in the specification that a fallback position has sufficient support. A search might help you have some insight as to what the examiner might find close enough to use a the base reference in a 103 rejection, although they often come up with something not found in your search. We all know about hindsight bias. Applicant arguments will often contain an assertion that the only reason it might seem obvious is that one has read the present application. This sometimes helps.
What questions does the examiner ask himself when determining obviousness?
They want to find the fewest number of references that, together, contain all of the limitations in your claim and that are arguably in the same field. They do not need to think their argument is airtight to make the rejection. They just make it and then it is your job to find the flaw in reasoning.
Is the obviousness issue not a concern in a provisional patent application?
As you will read every time someone uses the term "provisional patent", there is no such thing as a provisional patent. It is a provisional application for a patent. Obviousness is a ground of rejecting a claim. Provisional applications, unless directly converted to a non-provisional application (almost never done - sorry for being too detailed), need not have claims, are never examined, and therefore never subject to any form of rejection. However, the subject matter you end up relying upon to argue against obviousness or to allow amending to get out from a 103 rejection, should find adequate support in your provisional application.
Who decides "ordinary person skilled in the art" and how?
An office action almost never takes official notice of the qualification of the POSITA. They just assert that such an individual would have thought to combine the features of the references to solve the problem. A response to a 103 rejection could argue that the hypothetical person inventing in this field would have a high school education and that it would take a PhD, in some narrow field, of ordinary skill to see this as an obvious invention. I have not seen that and doubt if it would work. I do know that it often comes up in court and can make a difference there.
If you are planning on drafting this yourself you might consider looking up a recent patent in your field in the Public PAIR system of the USPTO. All of the back and forth between the examiner and the applicant is there for the public to study. You will see the original application, the grounds for rejection, and the examiner's responses. I understand that, on average, there are about 2.2 rejections before either there is an allowance or the applicant gives up. Studying a few cases from start to finish might help you draft something that anticipates what treatment it will receive.