If client actions take place locally, they are too early by a halfping.
If client actions take place on server, they are too late by a halfping.
There is a virtual 'halfping' point, a client estimate of server time:
halfping_time = (client_time+ping/2)
Maybe local client actions should be queued on the client (as well as sent to server) and performed at this halfping point?
Decided not to implement for sectors yet. Sectors influence movement (movement is actual client-time, so there be a conflict)
r1214 had an accidentally comitted test implementation, this was removed by r1215.