Yep. I see this too at work. InnoDB is in my opinion really good at handling concurrent workloads. So good I was surprised when I eventually found a project that was having locking issues. SHOW ENGINE INNODB STATUS showed queries had been waiting for hours on some locks they would never get. Yeah, it's a large and busy database, but it took me by surprise nevertheless.

It turns out, while InnoDB handles concurrent UPDATEs very efficiently, a combination of transactions that DELETE and INSERT rows - even just in the same general area of a table - will make the transactions wait for each other. Hence a workload that does a lot of inserts and deletes may get you into trouble. The solution is to change to READ COMMITTED or even READ UNCOMMITTED mode.

Thanks to Aaron from Ideeli for writing a proper blog about this, so I don't have to :-)

Update: Tagged Drizzle. Since this is an InnoDB issue, it of course applies just the same for Drizzle users too.