But WAR isn't really a counting stat -- you can (and many Twins regularly do ) get negative WAR. Danny Santana is at +4.7 offensive runs now, but if his next 140 PA are like Escobar's last 140 PA, that will probably drop precipitously (Escobar is down to -3.5 offensive runs right now).
And defensive WAR is even worse, I think, for small samples. dWAR leaderboards are littered with part-time players and even pitchers.
I see what you're saying. WAR isn't a counting stat because it can count backwards, but it's a cumulative stat in that a player needs to play a lot of innings defensively or acquire a lot of PAs to move the needle on their dWAR or oWAR, respectively.
In other words, Fuld will need to start being a sub-replacement level player for him to lose any WAR. Being replacement level offensively might mean a drop in average from here on out, and replacement level defensively might mean lower UZR/150, but replacement level will keep his WAR exactly where it is right now because nothing is being added or subtracted to it. That said, replacement level will not allow him to become the #1 CF in MLB
dWAR is finicky because it uses UZR, which is highly variable for small sample sizes (at least compared to oWAR). From what I can tell, this variability lends itself to assign defensive values that are so unrealistic that they override WAR's ability to be self-sample-size controlled.
So just how I read it is that a good (or lucky) fielder can increase (or decrease) their dWAR much more quickly with a series of improbable catches than a good (or lucky) batter can increase their oWAR with a series of improbable hits. For example: Chris Colabello has mashed with a series of doubles and homers since his callup to the tune of 3.8 batting runs. Fuld has gone full insane on defense over the last 30 days for 5.2 fielding runs. Next closest are Florimon and Nunez at 0.7. That's crazy.
*Should note that I'm mostly using fWAR for all of this, but mostly should apply to rWAR too.