There already are a number of quality answers to the questions raised, though there are a few more things that it might be useful to add.
Why use impedance controlled services? It has already been stated that this puts the responsibility for getting it right in the hands of the PCB house, but think for a moment what exactly that means. You want 50 Ohms, but the traces on your board may not be right. There are many possible reasons for this, thickness of materials (before pressing and after pressing are not the same, neither is the Dk value), the real influence of the solder resist is often an approximation and can be quite far from the figures used for calculating the impedance. Since some of these factors may not be apparent, when you ask for controlled impedance you are actually giving permission for the PCB manufacturer to adjust the trace widths as required to get the final impedance within usually 10% of the target value. To make this possible the PCB manufacturer needs to know which traces are the controlled impedance traces and the easiest way to show this is to make your controlled impedance traces to have a width that you do not use elsewhere. If you normally route your PCBs with 100um or 150um traces, and you find that the ideal width for 50 Ohms is 145.38um, then route these traces with 145um and then in the data accompanying the PCB design files, you state that 145um traces are 50 Ohms the PCB supplier now knows which are the traces to optimise their widths to give 50 Ohms on the finished PCB. You can do similar things when identifying controlled impedance differential pairs, but sometimes referring to both the trace width and the gap helps.
Does the 1/10th wavelength rule apply to regions with an impedance mismatch? The simple answer is not necessarily. Consider the two extremes: in the case of an open circuit, the whole of the energy reflects from this point feature doubling the voltage on the trace; or in the case of a short circuit, the whole of the energy reflects from this point feature reducing the voltage to zero. Thus length in these cases is not relevant. Whenever you have an impedance mismatch you always get some of the energy reflecting back towards the source (which may not appreciate this) and therefore you have increased path loss above that of a perfect microstrip feature. When this is at a very short distance from the source (say less than 1/10th wavelength) the reflections return to the driver sufficiently quickly that the drive may alter a little in response and so go some way towards hiding the discontinuity, as long as the discontinuity is not too large. Hence the rule of thumb that says if you have to have a discontinuity, the worst place you can put it is in the middle of the trace, and the best place is at an end. Thus the policy should be to try and get the impedance right, but if it is imperfect, it is less of an issue close to the end of the trace.
Is impedance control for production only or should it be prototype and production? If you can tolerate less than perfect operation of the prototype you might be able to justify making it a production only thing, however if you are going to measure anything about this RF link, you must have consistency. For example if you have a PCB mounted antenna and you need a matching network, there is no point performing the antenna impedance matching on a board without controlling the impedance of the board, or the first production build is going to come in different and will need matching again to be sure that the matching components are right for the new laminate. Similarly RF performance may be different between the two designs, so qualification testing on the prototype is not representative of the final product. What matters is consistency, so if you really want 50 Ohms, specify it as 50 Ohms for the prototype and also for production, and don’t change things when going from prototype to production.
Final remark, the impedance of the trace is a frequency dependent parameter. If you don’t say what frequency you want to use you will probably find that the PCB manufacturer will test at 100 MHz or perhaps 1 GHz. Between these frequencies there will be impedance differences, they may not be large, but they will exist. This means you may have to work with your PCB supplier to ensure that the cumulative tolerances don’t stray too far from what you need for your product.