Perhaps the LENR community could decide, precisely, which F&P style experiments are replicable and how they can be replicated? There does seem common agreement that positive LENR effects are genuinely replicable.
The variability here will not help the google guys.
For example: if an electrolysis experiment needs 10 cells to be tested of which on average 1 or 2 actually show excess heat, that must be understood by those replicating but it is fine.
Ascoli is right to warn that low success rates leave open the possibility of some error that induces the apparent success. But that is not a problem in a replication with good methodology where this is expected. For example, pre-screening could be done for active cells, followed by rigorous tests.
If however the situation is that the best LENR experiments, when replicated, may not generate anomalous results, even after best efforts to get this and multiple cells tested with best material selection and preparation, that is a problem.
I think it would be helpful to give google an explanation of why (perhaps) their tests so far have failed and a proscription for what needs to be done. Also helpful to find whether they have tested enough cells to deal with the 1 in 10 issue.
If high D loading is the key issue (I note that this is not universally agreed, though many think it) they should be told what methodology attains high loading. If they follow this to the letter and do not get it can i suggest that it may be that those who have claimed high D loading have in fcat suffered from the same mistakes that google identify in their paper? Perhaps google's attempts are as good as the best in the field so far. Anyway, this is one question that surely can be answered and should be answered for the google work to have significance for the LENR community.
There are many criterias to be fulfilled for F&P type LENR, many identified, and some criterias not yet fully identified and understood. That is why the sucessrate was low.
But again, to my knowledge the SPAWAR co-deposition studies gave better LENR reproducibility than F&P style and if google havent tried this allready, they absolutely should.