Cyton intermittently stops updating in OpenBCI GUI and drops from 250 Hz to about 62.5 Hz

Hi,
I’m having an intermittent issue with my Cyton on macOS that first showed up in the OpenBCI GUI.
Sometimes the GUI works normally, but other times I get errors such as "Data from board doesn’t change", or the GUI fails to start the session altogether. Disconnecting and reconnecting, or resetting the USB dongle, sometimes gets it working again, but the problem eventually comes back.
Because of this, I started testing the Cyton directly with BrainFlow to see what was actually being received from the board.

I’m using BoardIds.CYTON_BOARD with the normal Cyton USB dongle and serial connection.
I record for 10 seconds and count the number of samples returned by BrainFlow.
When everything is working normally, I get approximately 2484 to 2485 samples in 10 seconds, with mean timestamp spacing around 0.004 seconds and a rate around 250.3 Hz.

The packet numbers are consecutive with essentially no missing packets.

However, when the system enters the bad state, the exact same script gives only about 617 to 618 samples in 10 seconds, with mean timestamp spacing around 0.0161 seconds and a rate around 62.1 Hz.

For example, one bad run gave:

Samples received: 617
Mean timestamp spacing: 0.0161045 s
Median timestamp spacing: 0.0169806 s
Rate from timestamps: 62.094 Hz
Normal consecutive packet jumps: 616
Abnormal packet jumps: 0

So the packet counter is still perfectly consecutive. It does not look like I’m simply dropping packets.

The strange part is that it can switch between the two states. Earlier runs with the same board, Mac, serial port, and BrainFlow script produced the correct approximately 250 Hz rate, while other runs produced approximately 62 Hz.

I have also occasionally seen BrainFlow fail to initialize with "board doesnt send welcome characters" and "BOARD_NOT_READY_ERROR:7", before working again on a later attempt.

I found older forum discussions explaining that USB or OS buffering can make timestamps appear irregular even though the Cyton itself samples steadily. However, this seems different because in the bad state I am actually receiving only about 617 samples during a 10-second recording instead of about 2500.

Also, the bad rate is suspiciously close to 250 divided by 4, which is 62.5 Hz, and the corresponding interval is about 16 ms.

Has anyone seen this combination before?

The OpenBCI GUI sometimes reports that data from the board is not changing or fails to start.

BrainFlow sometimes receives the normal approximately 250 Hz.

Other times BrainFlow receives only approximately 62.5 Hz.

Packet numbers remain consecutive in both cases.

Any suggestions for what I should test next would be greatly appreciated