Does mixing pyplot state-machine and Object-Oriented API cause figure reference conflicts?

Direct answer

No. Calling a method on an Axes returned by plt.subplots() does not update pyplot's current-axes pointer. The behavior is deterministic: plt.plot() resolves its target as gcf().gca(), meaning it draws on the current figure's current axes—the last axes activated by a pyplot call, not the last axes you manipulated through an explicit reference.

Confirmed behavior vs. underlying mechanism

Confirmed across recent Matplotlib versions:

• plt.subplots() is itself a pyplot call, so at creation time it makes the new Figure current and its Axes current.
• Later OO calls on that Axes (ax.plot, ax.set_xlim) do not mutate pyplot's global state.
• plt.plot() after an OO call on a different figure targets the pyplot-current axes, not the OO-referenced one.

Likely explanation: pyplot maintains a global current-figure and current-axes pointer. Only pyplot entry points—plt.figure, plt.subplot, plt.subplots, plt.axes, plt.sca, and gca when it auto-creates—mutate that pointer. Explicit Axes methods operate on the object directly without side effects on the global state.

Minimal reproduction

import matplotlib.pyplot as plt

fig1, ax1 = plt.subplots()
fig2, ax2 = plt.subplots()

ax1.set_title("oo")       # explicit OO call
plt.plot([0, 1], [0, 1])  # pyplot call

Result: the title lands on ax1, but the line lands on ax2.

Fix and verification

To force pyplot to target an OO-held Axes, re-activate it explicitly:

plt.sca(ax1)              # or plt.figure(fig1); plt.sca(ax1)
plt.plot([0, 1], [0, 1])

Verify with a non-interactive backend:

import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt

fig1, ax1 = plt.subplots()
fig2, ax2 = plt.subplots()
ax1.set_title("oo")
plt.plot([0, 1], [0, 1])

assert plt.gca() is ax2
assert len(ax1.lines) == 0
assert len(ax2.lines) == 1

Recommendation and one diagnostic

Use one interface per figure. If you must mix them, call plt.sca(ax) immediately before any pyplot draw command. Avoid plt.gca() when you expect a specific Axes, because it can silently create a new Axes on the current Figure.

One missing detail changes the advice: if your Figure was created directly with matplotlib.figure.Figure() instead of plt.subplots(), it is not registered with pyplot's manager, so plt.gcf() and plt.plot() will not see it. Confirm the creation path if the behavior differs from the above.