Just to summarize the numbers below for future reference:

> let timeToPoll = 11 * 1e-6 -- from criterion
> let bytesPerPoll = 36802144/100000 -- from weigh
> let pollsPerSec = 1/timeToPoll
> let bytesPerSecond = bytesPerPoll * pollsPerSec
> pollsPerSec
90909.09090909091
> printf "%f" (pollsPerSec * bytesPerPoll)
33456494.545454547

So that's 90,909 polls and 33.5MB per second of allocations. That's
consistent with the below findings. The GC count for 100,000
polls (36.8MB) is 35 GCs, so that's roughly 1 GC per 1MB which
corresponds to GHC's default GC settings.

Before:

Case                Allocated  GCs
pollEvents 10           4,632    0
pollEvents 100         38,416    0
pollEvents 1000       369,112    0
pollEvents 10000    3,680,088    3
pollEvents 100000  36,802,144   35

benchmarking pollEvents/orig
time                 11.43 μs   (11.37 μs .. 11.49 μs)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 11.41 μs   (11.37 μs .. 11.48 μs)
std dev              167.3 ns   (122.4 ns .. 274.7 ns)
variance introduced by outliers: 11% (moderately inflated)

After

Case               Allocated  GCs
pollEvents 10          1,400    0
pollEvents 100         1,400    0
pollEvents 1000        1,400    0
pollEvents 10000       1,400    0
pollEvents 100000      1,400    0

benchmarking pollEvents/check
time                 11.05 μs   (11.00 μs .. 11.11 μs)
                     1.000 R²   (0.999 R² .. 1.000 R²)
mean                 11.04 μs   (10.99 μs .. 11.13 μs)
std dev              216.5 ns   (154.8 ns .. 296.8 ns)
variance introduced by outliers: 19% (moderately inflated)