I am trying to solve a 2D dynamic linear elasticity model parallel in time using Xbraid. The spatial domain is [0,1]x[0,1] and time domain [0,1]. For time integration I am using a backward Euler scheme and FEM for spatial descretization. I tried to run a simulation for a small spatial problem(4x4 quadrilateal elements) and 10000 time-steps and I noticed a weird behaviour of Xbraid's residual(a flactuation). At the beginning the residual decreases, for example starts from 3.24e-01 decreases for the next 30 iterations until 2.45e-04 .Then for the next 40 iterations there is a small increase until 2.09e-03 and the decreases again. Then the residual is decreasing normally until convergence. What may cause this behaviour?Is this normal? I want to note that I am using a standard coarsening and a FCF relaxation at all levels.V-cycle is used.