FlushCache messages might not be an actual IO stall
At a client site recently a SQL Server 2012 (Enterprise, SP1 + CU#9) server with 512GB RAM and 80 cores (40 cores assigned to SQL Server) that had been patched and restarted displayed the following symptoms after an index reorg operation was restarted: CPU saturation, with all cores available to SQL Server (but not all cores on the box) almost totally maxed out and the perfmon data showing an average of 96% CPU usage for those CPUs.