This appears to be a common problem when connecting ready-to-go SPI modules to the ESP8266. Many SPI modules, at least with the 3 out of the 4 MAX31855K peripheral boards I looked at, have a 10KΩ pull-up resistor on SPI Chip-Select. For most Arduino boards this may be fine; however, on the ESP8266 when using HSPI, pin GPIO15 is the Hardware SS (Subordinate Select). When Chip-Select from these peripheral boards is connected to GPIO15, the ESP8266, in most cases, will not boot. The GPIO15 pin serves a special purpose at boot time. To boot off the flash, GPIO15 must be held low or the device may start-up in an alternate boot mode. Most ESP8266 Development boards have a ~10K pull-down resistor on this pin. Combining the pull-down resistor with a pull-up resistor will results in an undefined logic level at boot time.

Some possible corrections:

• Select a different GPIO pin for SS and do a Soft Chip-Select
• Or remove the Chip-Select pull-up resistor from the peripheral board
• Or add an at boot isolation circuit to GPIO15. In Espressif's Hardware Design Guideline, Figure 1-10b. ESP8266EX UART SWAP they provide an example circuit to deal with a similar issue that can occur with MCUs connected to GPIO15. When reading the relevant first paragraph after Figure 1-10b, replace text MCU_RXD with SS. In addition to their text: GPIO5 or GPIO4 has to be used to control Q1 because at reset weak pull-ups are applied to most of the other GPIO pins. In the undriven state, Q1 is biased to be turned off. Weak pull-ups would interfere with this.
• Or lastly [Not recommended] decrease the pull-down resistor between GPIO15 and GND to 4.7KΩ. I think success with this option is dependent on the size of the pull-up resistor you are trying to counter, the amount of noise level on your board, and where the ESP8266 logic threshold is sitting. With 3.3V divided by a 4.7KΩ resistor and a typical 10KΩ pull-up resistor on the SPI module you arrive at 1.06V, which is a little high for a logic 0, but not too far from 0.8V CMOS V_IL. It may work some or most of the time; however, for proper noise handling, you want to be at a proper 0.4V CMOS V_OL. I don't like this solution. It is not a robust solution; however, I mention it here because you may see it used elsewhere.
  
  V_IL is the maximum input voltage that is guaranteed to be recognized as a logic 0. V_OL is the maximum driver output voltage for a logic 0, worst case for an output driver. The difference in CMOS V_IL and V_OL represents the amount of noise you can tolerate before your logic level enters the undefined region. Also interestingly if you use a 30KΩ pull-up against a 4.7KΩ pull-down you arrive at 0.45V. And 30KΩ just happens to be the minimum value you may see from the ESP8266's internal weak pull-up resistor. So if you have a pull-up of 30KΩ this solution could work; however, all the pull-ups I have seen in schematics have been 10 - 12KΩ at most.
  
  The Adafruit HUZZAH ESP8266 module just happens to use this 4.7KΩ value for GPIO15 pull-down. I think their module may do better than other ESP8266 Modules using 10-12KΩ.