fastResultLoading sometimes showing unrelated results, missing real ones

Created by: rvantonder

So I thought I was seeing weird highlighting issues before when using search, like, odd things being highlighted that shouldn't be, but this only happened once or twice and when I refreshed the issue disappeared. Turns out I am not crazy and can reproduce this reliably. This issue disappears when I put "enableFastResultLoading": false in my user settings so it is probably a regression in the changes there. Basically, stale/previously highlighted things show up, try this:

Run this on Sourcegraph.com: repo:^github\.com/ethereum/solidity$
Search for HAVE_z in this repository. You will see expected results, there's are hits for HAVE_Z3
Now change the pattern to HAVE_c and search. This will very likely show you a highlighted result for HAVE_Z which is not what we searched for, and no result for HAVE_C in that file (but there is a HAVE_C match in that file libsmtutil/SMTPortfolio.cpp).
Refresh, and you will see the expected, but previously missing HAVE_C highlighted and the HAVE_Z disappear.

In that particular file, there is a HAVE_Z and a HAVE_C close to each other, within a couple of lines of each other, so I suspect some calculation is messing up, but not sure why it fixes on refresh.

Trying to reproduce this issue, I also had very wacky stuff happening like entire file contents for results (again, some calculation issue going on) but I can't reproduce it reliably. Basically, I guarantee if you go to the above repo and alternate searching for HAVE_z and HAVE_c a couple of times you will see weird things. It disappears if I disable the fast loading flag.

Example: (Edit: the video does HAVE_c first then HAVE_z, I don't think this matters, just try alternate searches for these)

https://user-images.githubusercontent.com/888624/104407171-3e727380-551e-11eb-926a-32bfd3f4f56d.mp4