$ emacs -Q

;; Considering the next commands :

(defun pils/cycle-buffer-of-major-mode (&optional arg)
  "Switch to previous buffer of this major mode.
With ARG as \\[universal-argument], switch to the next instead."
  (interactive "P")
  (let ((switch-to-prev-buffer-skip
         (lambda (window buffer _bury-or-kill)
           (not (eq (buffer-local-value 'major-mode (window-buffer window))
                    (buffer-local-value 'major-mode buffer))))))
    (if arg (switch-to-next-buffer) (switch-to-prev-buffer nil 'append)))
  (when (eq (current-buffer) (window-old-buffer))
    (user-error "No other %s buffer available." major-mode)))

(defun pils/next-buffer-of-major-mode ()
  "Switch to the next buffer of this major mode."
  (interactive)
  (pils/cycle-buffer-of-major-mode t))

;; That we could temporary bind to :

(global-set-key (kbd "M-p") #'pils/cycle-buffer-of-major-mode)
(global-set-key (kbd "M-n") #'pils/next-buffer-of-major-mode)

Now gently but firmly, play theses emacs chords.

What happened ? The first command will put you in another mode.
That is unexpected ... Worse the second one will quickly jam.

The current implementation 'switch-to-prev-buffer-skip' fallback to the
first skipped buffer if no one have satisfied its anti-predicate. That's
why you could end up in another mode despite setting
'switch-to-prev-buffer-skip' to not select others modes.

Conservatively I fix that by checking if 'switch-to-prev-buffer-skip'
is a function.

The second issue is a bug in the implementation of
'switch-to-next-buffer'. This command should never return the same
buffer, it is wrote in its docstring, and it escaped me in commit
d0c7d8bc22.

So here a little patch to fix both.