Yet another BidiRow-related assertion failure

https://bugs.launchpad.net/ubuntu/+source/vte2.91/+bug/2011727 says:

VTE:ERROR:../src/ringview.cc:299:const vte::base::BidiRow* vte::base::RingView::get_bidirow(vte::grid::row_t) const: assertion failed (row >= m_start): (10011 >= 14401)

Most likely related to #2577 (closed), #2632 (closed), #2636 (closed). However, the actual vs. expected values differing by as much as 4390 is worrisome, and suggests a different underlying cause.

In the spirit of #2636 (comment 1784484) and #2636 (comment 1787869), I find it likely that the RingView incorrectly remains marked as up-to-date in a certain circumstance; a circumstance that can easily cause scroll_delta to jump by thousands.

What could this be?

• A rewrapping resize, renumbering the lines?
• A change in the scrollback buffer size?
• The escape sequence that wipes out the scrollback? (Probably not, since I think any input invalidates the RingView.)
• A switch from/to normal vs. alternate screen (each having their own row numbering)?
• The "Reset and Clear" menu entry (including possibly switching back from the alternate to the normal screen)?
• ... (anything else???)

All these should be revised.

It's possible that #2577 (closed) covers up this issue, but then it still results in slightly incorrect behavior (e.g. a click not handled properly) rather than a crash.