from a snapshot/event file, it is not easy to analyze in a quicky
the most common problem is different applications - multithread -
doing delete - select
not easy to say what is exactly wrong
- execute program in 1 process
- if delete - select.. are executed in same program
- try todo commit whenever possible
- for select use : select .. for update .. in that case a u-lock will be requested instead of s-lock and has to wait on other u-lock