The long-term (over more than one decade) X-ray emission from two massive stellar systems known to be particle accelerators is investigated using XMM-Newton. Their X-ray properties are interpreted taking into account recent information about their multiplicity and orbital parameters. The two targets, HD 168112 and HD 167971, appear to be overluminous in X-rays, lending additional support to the idea that a significant contribution of the X-ray emission comes from colliding-wind regions. The variability of the X-ray flux from HD 168112 is interpreted in terms of varying separation expected to follow the 1/D rule for adiabatic shocked winds. For HD 167971, marginal decrease of the X-ray flux in September 2002 could tentatively be explained by a partial wind eclipse in the close pair. No long-term variability could be demonstrated despite the significant difference of separation between 2002 and 2014. This suggests that the colliding-wind region in the wide orbit does not contribute a lot to the total X-ray emission, with a main contribution coming from the radiative shocked winds in the eclipsing pair. The latter result provides evidence that shocks in a colliding-wind region may be efficient particle accelerators even in the absence of bright X-ray emission, suggesting that particle acceleration may operate in a wide range of conditions. Finally, in hierarchical triple O-type systems, thermal X-rays do not necessarily constitute an efficient tracer to detect the wind-wind interaction in the long-period orbit.