esp-idf: RMT with TX loop enabled results in crash. Does not loop for 64 items.

EDIT: apologies - this may actually be two issues - if someone can advise perhaps I should split it into two?

When using RMT to generate a simple square wave, the application will crash if the rmt_config_t.tx_config.loop_en field is set to true and rmt_write_items() is called. If set to false it does not crash.

void rmt_task(void * pvParameter)
{
    rmt_config_t rmt_tx = {
        .rmt_mode = RMT_MODE_TX,
        .channel = RMT_CHANNEL_0,
        .gpio_num = GPIO_NUM_4,
        .mem_block_num = 1,
        .clk_div = 80,
        .tx_config.loop_en = true,  // <-- set this to 'true' for crash
        .tx_config.carrier_en = false,
        .tx_config.idle_level = RMT_IDLE_LEVEL_LOW,
        .tx_config.idle_output_en = true,
    };
    rmt_config(&rmt_tx);
    rmt_driver_install(rmt_tx.channel, 0, 0);

    // simple sequence of high for one clock, low for one clock
    rmt_item32_t items[2] = {0};
    items[0].level0 = 1;
    items[0].duration0 = 1;
    items[0].level1 = 0;
    items[0].duration1 = 1;

    // terminate sequence with a zero-duration item
    items[1].level0 = 0;
    items[1].duration0 = 0;

    // Note: I've also tried this with a single item - same result
    rmt_write_items(RMT_CHANNEL_0, items, 2, false);

    while(1)
        ;
}

void app_main()
{
    xTaskCreate(&rmt_task, "rmt_task", 2048, NULL, 5, NULL);
}

Sometimes when it crashes I see log messages such as:

W (90) boot: PRO CPU has been reset by WDT.

or

***ERROR*** A stack overflow in task ???D??@(?? has been detected.

Other times it just restarts with no error.

100% reproducible with my setup. I’m using ESP-IDF 27574a31e7c41f6e22e10e9b9d7713c6d49f795d