Random data loss, less than 0.1% of the time - how to prevent it

[Ted Byers]

I do not know even how to debug this, let alone how to fix, or prevent it. What I observe is that for, say, 20,000 SQL statements that insert data into my database, there is a random number, typically between 5 and 10, of statements that fail to insert the data. What is more puzzling, and perturbing, is that these failures happen within a very short, randomly occuring, period (typically less than 5 minutes), with an average of one a week. Sometimes there are weeks in which it doesn't happen and sometimes there are two in a week. I have never seen a week with 3 such periods. The tables involved get only INSERT SQL statements, and select statements (but the latter are obviously not a problem), and NEVER UPATE nor DELETES.

It is version 5.5, if that matters.

I guess I have two questions.

1) How does one begin to investigate a problem like this, when it happens so rarely and so randomly?

2) what can be done to guarantee that I never suffer this sort of data loss again? Is there something in the configuration of MySQL that can help, or will some kind of buffer (some sort of daemon or service that is running 24x7 and which takes the data and is responsible for passing it on to mysql) help?

I suppose a DBA would know the answers to these questions, but I am not a DBA, rather I am a programmer. Any help would be appreciated.

Thanks

Ted