ESXi host probes and iterate through all the paths before finding the active path and doing a path failover. ESXi host internally uses mode sense and RTPG requests to probe the paths and each of these request uses 10 seconds of time out. In case of SP reboot, ESXi host does not get response while probing the path from the target and waits for the commands to time out. For each path the time out is around 20 seconds (10 seconds time out for mode sense and RTPG respectively).
The overall delay is directly proportional to the number of paths per SP. For example: If you have 4 paths per SP, then overall fail over time will be double of 2 paths per SP. Hitting a long time out is unlikely because the SP will force its link down, which triggers Registered State Change Notifications (RSCN) across the fabric and ESXi host fails over immediately.
Note: The impact is seen only if there are active IO in progress during SP reboot.