excessive CPU time spent on gc (even after manually adjusting gc thresholds)

During the process I'm running I very quickly get this warning (for pretty much each worker)
```
distributed.utils_perf - WARNING - full garbage collections took 36% CPU time recently (threshold: 10%)
distributed.utils_perf - WARNING - full garbage collections took 34% CPU time recently (threshold: 10%)
distributed.utils_perf - WARNING - full garbage collections took 35% CPU time recently (threshold: 10%)
distributed.utils_perf - WARNING - full garbage collections took 34% CPU time recently (threshold: 10%)
distributed.utils_perf - WARNING - full garbage collections took 35% CPU time recently (threshold: 10%)
distributed.utils_perf - WARNING - full garbage collections took 35% CPU time recently (threshold: 10%)
```
However, the percent of memory used is by each worker is low -

I have plenty of memory to work with so my question is - is it possible dask is aggressively doing gc? If so is it possible to change the gc threshold/collect less aggressively?

I've tried to manually adjust the `gc.threshold` but this seems to have no effect
```
g0, g1, g2 = gc.get_threshold()
gc.set_threshold(g0*10, g1*10, g2*10)
```

(using distributed 1.28.1 and dask 1.2.2)