Pruning to maximum edge size doesn't necessarily meet target.

[rwest]

A recent run, which was set to prune the model edge to a maximum of 20,000 species, had this edge size behaviour:

We had set the limit at 20,000 because we knew it would run out of memory around 40,000.
When it first exceeded 20,000, it was held around the 20k-25k mark for quite a while, but gradually got out of control.
Sure enough, the job ran out of memory and died, when it reached about 40,000. It seems that sometimes there are a lot of species that, for some reason or another, are not, or cannot be, pruned with the current implementation.

[DetlevCM]

Just because I see this: From my impression, the maximum edge is not a "hard limit". Setting RMG to prune very aggressively, RMG has often exceeded the target edge species count in my runs.
Spontaneously I would say the edge flux can override the maximum edge species setting.

[rwest]

@faribas and I have also seen this behaviour. I had a plot of edge size against iteration, showing the pruning fighting a losing battle to get it back down to the limit, but I can't find the file now. We didn't look into the cause, but did note that the size was not monotonically increasing (occasionally the pruning would get it back down again, just not far enough often enough).

[rwest]

Haha. The graph I was looking for is above. I feel so stupid sometimes.