mergefactor=1 questions

[Phillip Farber <pf...@umich.edu>]

In order to make maximal use of our storage by avoiding the dead 2x 
overhead needed to optimize the index we are considering setting 
mergefactor=1 and living with the slow indexing performance which is not 
a problem in our use case.

Some questions:

1) Does mergefactor=1 mean that the size of the index on disk increases 
only due to <add/>s or is there some sort of merging that happens that 
temporarily inflates disk usage?

2) It was mentioned that, with "per-segment readers", an optimized index 
may not be the best option.

What are per-segment readers? Is this configurable or some sort of default?

What are the cases where an optimized index (one segment) might not be 
the best option?

Thanks!

Phil