Long-format streaming breaks acquisition GUIs. LabRecorder, XDF and a custom loop against automatic staging and the drift of a wet montage.
A few-minute session reveals no problem. A full night, eight hours of continuous stream, exposes the hidden limits of the software chain. The GUI rarely gives out by crashing, more often through a gradual memory leak or accumulation in a buffer that eventually saturates.
Why long format breaks
An acquisition GUI often keeps in memory enough data to plot and inspect the signal. Over eight hours, that buffer grows without bound if nothing empties it. LabRecorder with an XDF format handles continuous writing better, but remains sensitive to disk space and I/O latency as a single file swells.
A custom acquisition loop, with no graphical rendering, shifts the constraint to disk writing, which is far more predictable.
Staging and the human limit
Automatic staging splits the recording into timestamped segments written then closed on the fly. Memory stays bounded and an incident takes out only one segment. But no software trick fixes the physical limit: a wet montage drifts after 2 to 3 h, the gel dries and impedance rises.
The takeaway: for long format, bound memory through staging and accept that the real limit is the electrode, not the software.