Here's a simple example to reproduce:
import numpy as np
import mne
fs = 500
n_chans = 10
data = np.random.randn(n_chans, fs * 100) * 5e-6
info = mne.create_info(n_chans, fs, "eeg")
raw = mne.io.RawArray(data, info)
raw.annotations.append(onset=[1, 5, 9, 12, 15, 19],
duration=[2, 5, 3, 3, 5, 3],
description=["A1", "A1", "B2", "C3", "A1", "B2"])
raw.plot(n_channels=n_chans)
Once the plot window is open, increase the time window and notice that annotations outside the initial time window are not drawn. Here's what the plot looks like after increasing the window 13 times (notice that annotations from 10s are not drawn, but the overview bar still shows them):

@drammock do you have time to take a look?
Here's a simple example to reproduce:
Once the plot window is open, increase the time window and notice that annotations outside the initial time window are not drawn. Here's what the plot looks like after increasing the window 13 times (notice that annotations from 10s are not drawn, but the overview bar still shows them):
@drammock do you have time to take a look?