The cpuWaitPct counter in our MP is calculated using the cpuWait metric from vCenter. cpuWait is in milliseconds; we convert this to percentage by simply calculating the cpuWait in ms over the performance sample interval in ms.
So cpuWaitPct is definitely not simply an inverse of cpuUsedPct - although if you display both on the same graph you can see what looks like inverse behaviour, this is due to the nature of the metrics (as a VM uses CPU, it is of course not waiting for CPU...)
We do deliver cpuWaitPct as an average across all vCPU cores, as with cpuUsedPct.
And we don't deliver Idle for vCPUs in the current release - however it will be added in our next update, as cpuIdlePct.
I'd also advise adding cpuReady into your troubleshooting, we capture this as cpuReadyPct. In 5.7 release we added cpuSwapWait and cpuLatency, also important metrics.
BTW - apologies for delay in replying, this forum does not always email me for new posts. And moving it to the Cloud hasn't helped
Hope the above helps!