dude pls...do you expect an ANSI norm for datafeeds? If you want to theorize, at least do it right.
Your pompous lambda (I don't know who do you want to impress with it) can be exactly measured by the amount of ticks that you receive from the third party data feed vs. the ticks you receive directly from the exchange. Even uptime is accounted for since...well you don't receive ticks during downtime.
1000 ticks on exchange/your datafeed delivers 800 ticks -> not so reliable
1000 ticks on exchange/your datafeed delivers 900 ticks -> a bit more reliable
So if you still want to call your failure rate lambda, you could say that in the first case your lambda is 20% and in the second case 10%.
There are only two qualifiers for a datafeed: completeness and speed. Speed is not the question here, although you could compare timestamps and include it in your quest for reliability. And if you now come up with technicals like jitter and lag, these are NOT qualifiers for a datafeed but for connectivity.
So please think before you post again, the hole you dug yourself into is already deep enough
Grade F.