Different file size limits for fif files in Neuromag and mne

Was it intentional to set the file size limit to precisely 2gb? It seems that the Neuromag Vectorview uses a limit of 2gb minus 100mb (although this is configurable, see page 54 of the user manual). No clue whether this is common for other Neuromag's scanners.

I'm wondering because some of my recordings happen to be just below 2gb, so that the MEG scanner split the files, but when running the data through mne-bids, I ended up with a single file. Not that this is a big problem, but when I tried to figure out based on a raw instance whether a file would be split once saved, I relied on the raw.filenames attribute. In this particular edge case, that approach doesn't work.

Edit: That being said, is there a way to predict how large a raw object will be after saving on disk?

Start by tracing how FIF recordings are split during saving and how mne-bids uses the raw.filenames attribute. Compare the 2 GB limit with the Neuromag Vectorview behavior described in the issue, and determine whether saved size can be predicted reliably. Done should clarify the intended limit and the behavior of raw.filenames for split recordings.