AVR Freaks

Actually I try to set up a TWI slave with the ATmega328PB (2 TWI ports).

The master is running with Peter Fleurys I2C library on a ATmega328P (1 TWI port).
With the master I'm checking, if the TWI slave is available on a specific address:

/*************************************************************************
  Issues a start condition and sends address and transfer direction.
  return 0 = device accessible, 1= failed to access device
*************************************************************************/
unsigned char i2c_start(unsigned char address)
{
    uint8_t   twst;

	// send START condition
	TWCR = (1<<TWINT) | (1<<TWSTA) | (1<<TWEN);

	// wait until transmission completed
	while(!(TWCR & (1<<TWINT)));

	// check value of TWI Status Register. Mask prescaler bits.
	twst = TW_STATUS & 0xF8;
	if ( (twst != TW_START) && (twst != TW_REP_START)) return 1;

	// send device address
	TWDR = address;
	TWCR = (1<<TWINT) | (1<<TWEN);

	// wail until transmission completed and ACK/NACK has been received
	while(!(TWCR & (1<<TWINT)));

	// check value of TWI Status Register. Mask prescaler bits.
	twst = TW_STATUS & 0xF8;
	if ( (twst != TW_MT_SLA_ACK) && (twst != TW_MR_SLA_ACK) ) return 1;

	return 0;

}/* i2c_start */

This works fine, as you can see on the picture of the scope (attached).
Now the problem is, that the slave should give an acknowledge, and, as I understand the procedure, should release the SDA-line level back to high (green line).
But SDA is still beeing pulled low by the slave. If I switch off the slave mcu (SDA released), then suddenly the master reports that the slave would be available...

The slave is initialized this way:

void init_twi_slave(uint8_t adr)
{
    TWAR0= adr << 1; // set twi slave address (skip over TWGCE bit)
	TWCR0 &= ~(1<<TWSTA)|(1<<TWSTO);
	TWCR0|= (1<<TWEA) | (1<<TWEN)|(1<<TWIE);
	buffer_adr=0xFF;
	sei();
}

When the master is requesting the slave, the ISR of the slave comes up with the TW_STATUS = 0xA8 (..I send TW_STATUS over UART to check)

ISR (TWI0_vect)
{ ....
...

     case TW_ST_SLA_ACK:      // 0xA8: SLA+R received, ACK returned
     case TW_ST_DATA_ACK:     // 0xB8: data transmitted, ACK received
       if (i2c_state == 1) {
	 i2c_slave_action(0); // Call Read I2C Action (rw_status = 0)

	 TWDR = regdata;      // Store data in TWDR register
	 i2c_state = 0;	      // Reset I2C State
       }	   	  

       TWCR |= (1<<TWINT);    // Clear TWINT Flag
       break;

     case TW_ST_DATA_NACK:    // 0xC0: data transmitted, NACK received
     case TW_ST_LAST_DATA:    // 0xC8: last data byte transmitted, ACK received
     case TW_BUS_ERROR:       // 0x00: illegal start or stop condition
     default:
       TWCR |= (1<<TWINT);    // Clear TWINT Flag
}

Doesn't  TW_STATUS = 0xA8 mean, that the mcu should have already released the SDA line?

Any ideas, how to fix this or where I'm going wrong?

Many Thanks!