Seems to me that it depends on whether it is a Base stock of Group I, II or III ranked base for the oil with Class I being fully synthetic and Class III permisible "natural" base - stock . My understanding is that it is when the base for the oil is Stage II or III then you get molecular breakdown. They start off with varying length molecules that decomp to a shorter length losing efficacy. Along with this is the extra loading and using up of the reagents that contribute to pour-point (that lower number) and the stabilisers for the upper temp shear-resistance.
These aren't my only selection parameters - also Flash Point, VI etc has a bearing but this is not about selecting an oil.
I think it is important to consider not only the API ratings but the ACEA ratings - will the oil "hold" structure in service for the intended interval? etc. These have far more meaning than they seem to be given credit for and until I wade through the body of work I am currently reading have been my primary (plus Chrysler recommendation) guidance for oil selection.
I am a proponent of by-pass oil filters and am currently working on trying to plan one into the Australian under-hood arrangement. A puzzle to say the least but well on the way to being solved

. That 10 % approx that is hived off for by-pass filtration will put no extra load on the pump as the full-flow filter will still be carrying 90% of flow. There is no question that this is attainable by the oil pump.
IF resistance is higher in the lines to the by-pass filter it weill be as a rsult of further flow restrictions and the default flow will be through the primary (full-flow) filter - it will merely restrict how much flow the by-pass unit gets will it not?
Thinner oils and greater flow on a viscosity rating that is measured at 100 deg C? - It will not going vary so much when it changes by 10 on THAT index. Surely VI (Viscosity Index) is more important in this context.
I come back to my point regarding base stock Groups and the relastionship to decomposition.
Cudos to by-pass filters.