Sleep mode current consumption fluctuating

Start with the DeepSleep::launch and DeepSleep::recover entry points in the provided deep_sleep.h-based code, and reproduce the fluctuation using the minimal sleep-loop program described. Compare the current draw across successive sleep_goto_sleep_until calls and check the Pico datasheet and the linked dormant-mode report. Done means identifying why consumption changes and documenting or implementing a reliable correction.

When I enter sleep mode (not dormant mode) the current consumption of my Pico is unreliable. I can successfully enter and exit sleep mode and I enable all necessary clocks I believe:

#include "deep_sleep.h"

uint scb_orig;
uint en0_orig;
uint en1_orig;

void DeepSleep::recover() {
  // Re-enable ring oscillator control
  rosc_write(&rosc_hw->ctrl, ROSC_CTRL_ENABLE_BITS);

  // reset procs back to default
  scb_hw->scr = scb_orig;
  clocks_hw->sleep_en0 = en0_orig;
  clocks_hw->sleep_en1 = en1_orig;

  // reset clocks
  clocks_init();

  set_sys_clock_khz(133000, true);
}

void DeepSleep::launch(int8_t minutes, int8_t seconds) {
  if (minutes < 0 || minutes > 59 || seconds < 0 || seconds > 59) {
    return;
  }
  // save current values for clocks
  scb_orig = scb_hw->scr;
  en0_orig = clocks_hw->sleep_en0;
  en1_orig = clocks_hw->sleep_en1;
  sleep_run_from_xosc();
  datetime_t t = {
    .year  = 2024,
    .month = 01,
    .day   = 27,
    .dotw  = 1,
    .hour  = 12,
    .min   = 00,
    .sec   = 00
  };

  datetime_t t_alarm = {
    .year  = 2024,
    .month = 01,
    .day   = 27,
    .dotw  = 1,
    .hour  = 12,
    .min   = minutes,
    .sec   = seconds
  };

  // Start the RTC
  rtc_init();
  rtc_set_datetime(&t);
  sleep_goto_sleep_until(&t_alarm, &recover);
}

I can then enter sleep mode by calling DeepSleep::launch(0, 10) in order to sleep for, in this case, 10 seconds.

Whenever I do this the current consumption is often as low as 1.3mA the first time I enter deep sleep, successive calls however often cause a higher current consumption of about 4mA. Does anyone know why? Its also really hard to tell when this problem occurs, I tested this on a Pico without any sensors connected. I also tried a super simple program where the Pico only goes into deep sleep over and over again, and it appears to happen less, but occasionally it still happens. And once the current draw is the higher one in sleep mode it keeps being this high, like it never returns to only drawing the 1.3mA in sleep mode, which is also really odd.

I found one related post that seems to be experiencing the same problem, albeit being in dormant mode: https://forums.raspberrypi.com/viewtopic.php?t=336273

Maybe theres some hardware features that aren't always turned off? I tried looking through the datasheet, but couldn't find a solution...

Help would be much appreciated!