Hi,
after a random number (usually around a few hundred thousand) of measurements, the sensor starts sending the value 64 and the DIAG_STATUS bit in the CONV_STATUS register is set to 1. I have replaced the sensor and the error still occurs.
Hardware setup:

Supply voltage: 3.3 V
I2C frequency: about 300 kHz
INT pins not connected (not used due to lack of free MCU pins)
No devices other than the two TMAG5273 sensors share the I2C line
TMAG5273 Configuration:
DEVICE_CONFIG_1 (0x2A):
I2C_RD 2h = 1-byte I2C read command for 8 bit sensor MSB data and conversion status
CONV_AVG 2h = 4x average, 3.1-kSPS (3-axes) or 8.0-kSPS (1 axis)
MAG_TEMPCO 1h = 0.12%/ deg C (NdBFe)
CRC_EN 0h = CRC disabled
SENSOR_CONFIG_1 (0x40):
MAG_CH_EN 4h = Z channel enabled
SLEEPTIME 0h = 1ms
SENSOR_CONFIG_2 (0x03):
X_Y_RANGE 1h = ±80mT (TMAG5273A1) or ±266mT (TMAG5273A2)
Z_RANGE 1h = ±80mT (TMAG5273A1) or ±266mT (TMAG5273A2)
All others fields in default state
INT_CONFIG_1 (0x01):
MASK_INTB 1h = INT pin is disabled (for wake-up and trigger functions)
All others fields in default state
All others registers in default state
Method of reading data:
1. Wake up, and trigger conversion on first sensor (DEVICE_CONFIG_2 = 0x00 (standby mode)):

then, the same for the second sensor:

2. Read data on both sensors:

3. Put both sensors to sleep (DEVICE_CONFIG_2 = 0x01 (sleep mode)):

Summary:
When reading data on one of the sensors (sometimes also on two), the value of the CONV_STATUS register is 0x03, i.e.:
RESULT_STATUS 1h = Conversion data complete
DIAG_STATUS 1h = Diag fail detected
Diag fail detected is described as
Detect any internal diagnostics fail which include VCC UV, internal
memory CRC error, INT pin error and internal clock error. Ignore this
bit status if VCC < 2.3V
I would like to ask for help as I don't know how to solve this problem.


