MAX3232E: why the capacitor value need to change according to VCC?

I ask this because one of my customer is using 5V VCC and 0.1uF cap. And they will get below waveforms after operating correctly a period of time:

what may cause such malfunction? How should I check?

Besides, the customer says that they can not observe below waveforms of the charge pump clearly, I'm not sure if our device will behave as below, our datasheet does not provide waveforms that shows how V+, V- looks like.

Howard,

The capacitor values can be tweaked to control the ripple of the output voltage. You can take a look at this post for some details (https://e2e.ti.com/support/interface/rs232/f/392/t/463809).

Can you describe what happened during the malfunction? Did the output Dout become zero during communication? Can you check to see if the V+ voltage is normal (above 5V)?

MAX3232E has similar power up ramp behavior as what's shown in the figure.

Regards,

Hao

Hao,

thank you.

We've checked the device's V+ and V-.

Bad device is:

Good device is 

All waveforms are obtained with no load. Any insight what might cause the device break?

Howard,
Thanks for your information. It looks like the V- side might have some loading issues. Can you please check the voltage rating of the capacitors? We recommend the capacitors with at least 10V voltage rating or even 16V due to the charging voltage of V+ and V-. If the device stops function, can you check if the communication could resume 1) by power cycling the device, 2) by replacing a new part on the same board, 3) by replacing C2/C4 cap on the board?
Regards,
Hao

Yes, the four capacitors' voltage rating is 16V.
The communication cannot resume by power cycling, or by replacing C2/C4 cap on board. The voltage of V- is still abnormal.
The communication can resume by replacing a new part on the same board.
It's mass production project, so it may be related to their working conditions.
The customer would like us to help determine the root cause.