Though PingER has very low network bandwidth requirements (e.g. 100bits/sec on average per monitor-remote host pair), it is not probe-less. At roughly regular intervals (30 mins) the monitors send up to 30 ICMP ping probes to each remote host until it they get 10 responses from each host or time out.
As currently implemented the remote hosts have fixed IP addresses, thus it would not work for mobile clients. One would be able to monitor from your monitoring hosts as far as the fixed IP address is available (e.g. to the client gateway/NAT box etc.) which might be good enough for what you need (if one is just interested in the performance of the network). Of course this assumes the client gateway/NAT box etc. has to respond to pings.
For mobile clients I suspect one needs to install an App. Before an App is installed the client owner has to believe it will provide some benefit to her/him and will not cause problems. Thus with Ookla SpeedTest and MobiPerf the client is provided with an estimate of network performance. At the same time the provider can save the estimates and if GPS location is turned on, can know where the performance is available, hence providing lots of data for mining and plotting on maps (e.g. see the Akamai app). Similarly downloading of data (say apps) from a location (say an apps store) gives information on performance to the app store owner.
There are several mobile ping clients/apps available such as ping lite, pingtool, network analyzer etc. I imagine they can capture the data and may use it for mining. I suspect they could be improved upon to make them more attractive (e.g. give extra information on say Mean Opinion Score (MOS) and how well VoIP is likely to work, the Directness of the connection (i.e. comparing the RTT to that expected on a great circle route at the speed of light), indications of the quality of the connection (expectations, derived from measured jitter, RTT, loss, or for derived throughput etc.), also combining with a visual traceroute etc. I am not clear this would be of interest to an ISP. Maybe a startup would be interested, or it could be added to existing apps, such as mentioned above.