Light Fixture, Fan and GFCI share supply, all function except Light Fixture -- unless power is consumed via GFCI

Question

I have concerning behavior from a light switch and GFCI outlet. I apologize for the length, I'm trying to be thorough as I have a pretty good understand of how wiring should work, but it's...not...So I'm missing something somewhere.

Setup

The junction of concern is supplied on the load side of an upstream GFCI[A] Outlet. The junction box has a GFCI[B] Outlet, Light Fixture and Bathroom Fan all pigtailed together from GFCI[A] load (i.e. GFCI[B] supply). I feel confident the issue is in this junction box, but don't let me convince you of anything yet.

Breaker -> GFCI[A] ==Load==> Junction Box [ GFCI[B] | Light Switch ==> Light Fixture | 3 Speed Bathroom Fan (no light) ]

There is no connection to ground; this is old home with two wire install on this circuit. The Light Fixture, Light Switch, and GFCI[B] are new work with Romex, and are all connected together via ground for posterity, but there is no connection to ground upstream at GFCI[A].

Behavior

In short, the light fixture will only turn on under two conditions

1. GFCI[B] outlet is removed (wires still attached to pigtail).
2. Power is consumed by some appliance via the GFCI[A] or GFCI[B] (with light switch in on position, of course)

By point 2. I mean I have an electric kettle with a switch on it at GFCI[B] and Toaster on GFCI[A]. With either plugged in and I flip power to the kettle/toaster, the light fixture comes on with it (if light switch is on, duh). The light fixture will remain on until I flip the light fixture switch to off.

Once the Light Fixture is on, it will remain on regardless of GFCI[A/B]

The Bathroom Fan functions at all times and GFCI[A/B] function at all times.

Setup in Light Switch

Single pole switch, with black wire power-in going to top terminal, and black wire power-out at bottom terminal.

Two white wires wire nutted together. Inspected for solid contact.

Ground wires wire nutted together, and pig-tail bonded to metal box. No ground upstream, just good habits.

Setup in GFCI[B] Junction Box

Remember this is all load side of GFCI[A]

Black wire (from load GFCI[A]) is wire-nutted to 3 other black wires. They connect to:

1. GFCI[B] LINE/SUPPLY terminal
2. Bathroom Fan
3. Light Switch ==> Light Fixture

4 White wires all are bonded via "IDEAL In Sure" wire connector (didn't have Wago's at my store). White wire from GFCI[B] is on LINE/SUPPLY terminal.

Results of testing and investigation

1. Black to White across GFCI[B] terminals reads 119v (in essence black pigtail to white pigtail reads 119v)
2. Black to Black across light switch open reads 119v (0v when closed) regardless if GFCI[B] is connected or not.
3. Plug-in outlet tester shows "Open Ground" on GFCI[B] and no other codes.
4. Disconnecting white/black wires from GFCI[B] allows Light Fixture to operate as expected.
5. Tripping GFCI[B] "test" does not allow Light Fixture to operate. (was trying to mimic it being disconnected)
6. Tripping GFCI[A] "test" removes power from everything downstream, as expected.
7. Bathroom Fan works at all times.
8. Testing Black/White to Ground reads ~60-80v, but since this is not actually connected to Ground I believe this is meaningless.
9. With GFCI[B] connected, and electric kettle pulling power from it, the Light Fixture will turn on. Turn off the electric kettle and Light Fixture remains on, it very slightly brightens.
10. If I flip the Light Switch to on (with GCI[B] installed), the light remains off until I remove either of Black or White wire connected to GFCI[B]. Honest to god just leaving the screw loose, and then wiggling the wire until it looses contact with the outlet, then the light pops on (and stays on).
11. I originally thought this was a loose wire and maybe using the electric kettle caused a neutral to heat up, bend/expand, and then provide sufficient contact for the light to work. I've double triple double checked that wires are snug, and in contact.
12. Breaker is functioning: trips during a short.
13. Pulling power from the upstream GFCI[A] (toaster oven) will turn on the light, same as GFCI[B]
14. Pulling power from Bathroom Fan will not turn on the light
15. Best for last: With GFCI[B] installed (and light won't turn on) No-Contact Voltage tester reads no voltage with Light Switch off. With Light Switch on, the tester beeps indicating voltage is in there. So something gets through

My thoughts

Absolutely stumped.

The only thing left I can think to try is a new GFCI outlet, or just a regular outlet. A regular outlet should suffice since it is on the load of GFCI[A]. But I am not confident this solves whatever issue is going on. Perhaps GFCI[B] is faulty and has a small short? Enough to cause a voltage drop to make the light not turn on and the no-contact beep, but enough for the bathroom fan to operate. I read 119v across the light switch in this setup, however.

Nothing can be leaking through ground since there isn't ground.

To me the absolute strangest part is that removing GFCI[B] makes the light work. Or that the Bathroom Fan does work while it's connected.

I have no other ideas. Thank you for your time.

Edit: The light is working now, and I'm noticing it works more often during daytime than night time. It really suggests to me it's temperature dependent (winter here) and a loose wire, perhaps at A