For now, lha was judged right behind the interval checked was done, this behavior has some flaws:

Suppose we have multiple nodes A, B, C, D, the A node have the smallest elapsed from 1 check, glider choice A as winner node now. But in fact, node B(or C D) may be faster than node A most of the time.

This situation often occurs on routes from mainland China to overseas，Shorter the checkinterval and Bigger the checktolerance may improve the flaws but not perfect.

How about check continuously（let's call inner-check here） during checkinterval,then next checkinterval will use the average elapsed value as winner.

For checkinterval=3600 i mostly used, let inner-check=60, we will get the the very fair result from the past 1 hour.

For glider first start, old way is fine if we have no result cache to use.